[{"@search.score":1.0,"documentId":"9f910752-74dd-487e-96d8-75755ec72eb8","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Satya Nadella","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775599881893001JQFY","langLocale":"en-US","sessionId":"1775599881893001JQFY","sessionInstanceId":"1775599881893001JQFY","sessionCode":"KEY01","sessionCodeNormalized":"EN-US-KEY01","title":"Microsoft Build opening keynote","sortTitle":"microsoft build opening keynote","description":"Satya Nadella and Microsoft leaders share how Microsoft is creating new opportunity for developers across our platforms in this era of AI.\n\nThere is limited seating in the Festival Pavillion for the keynote. Please arrive early to secure a spot. Additional keynote viewing areas will be available in the outdoor theater, BATS Theater (building B), Cowell Theater, and inside Gateway Pavillion, level 2, at all theaters.","aiDescription":"<p><strong>Opening and Vision for Developers:</strong> The session begins with <a class=\"medius-aisummary\" data-starttime=\"00:01:23\">00:01:23</a> by greeting developers at Microsoft Build in San Francisco, emphasizing how developer conferences capture moments of technological transformation. The focus is on understanding the changing tech stack and the opportunities it brings for developers, companies, and society. The central idea is participating in the “frontier intelligence ecosystem” — not bound to one platform but about building compounding value on top of it. Satya Nadella lays out that the purpose of this conference is to unpack the emerging AI stack from compute infrastructure to agents, governance, and the developer tools that empower innovation.</p>\n\n<p><strong>Infrastructure and Edge Compute Innovations:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>, Nadella outlines the AI stack’s foundation—an interconnected compute fabric running from edge to cloud. He highlights the power of Windows at the edge, citing how PCs now host powerful local AI capabilities. Tools like Outlook’s summarization or PowerPoint’s visual aids run on local AI models using NPUs and GPUs. Microsoft expands Windows ML and Windows AI, introducing onboard reasoning models like Ion Instruct and agentic models such as Ion Plan. Hardware innovation is also emphasized, with collaboration from AMD, Intel, Qualcomm, and NVIDIA—the latter powering the new RTX Spark SoC and the Surface Laptop Ultra. Developers are introduced to the Surface RTX Spark Dev Box at <a class=\"medius-aisummary\" data-starttime=\"00:08:06\">00:08:06</a>, a machine delivering petaflop-level AI compute designed specifically for software creation. This evolution represents “unmetered intelligence” directly on user devices.</p>\n\n<p><strong>Developer Tools and Local AI Experiences:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:12:08\">00:12:08</a>, Kayla demonstrates how Windows becomes a distraction-free, AI-enhanced development environment running on the Surface RTX Spark. She shows off vertical taskbars, performance-optimized dev drives, new intelligent terminals integrated with GitHub Copilot, and WSL containers with full GPU utilization. The presentation details how developers can build and test sophisticated AI agents locally while maintaining productivity, privacy, and cost efficiency. This part of the keynote reinforces that Windows now offers the richest native development experience with Linux and Mac ecosystem tools—merging versatility, security, and multi-model AI capabilities directly on the developer desktop.</p>\n\n<p><strong>Cloud Scale and Responsible Infrastructure:</strong> As the focus shifts to the cloud at <a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>, Nadella discusses Azure’s rapid global expansion, scaling both capacity and environmental stewardship. Data centers are being engineered for efficiency—achieving near-zero water operation and sustainable power design. He unveils innovations like the AMD MI300 and the custom ARM-based Cobalt CPUs and networks designed for the dominant AI workloads: training, inference, and agent runtime. Collaborations with NVIDIA are outlined, emphasizing cloud architecture tuned for “tokens per dollar per watt.” Nadella and Jensen Huang jointly discuss the synergy between AI PCs (RTX Spark) and advanced data centers powered by systems like Vera Rubin, which enable secure confidential computing and large-scale agent workloads—all forming the foundation of the AI-driven ecosystem of the future.</p>\n\n<p><strong>Agentic Devices and Project Solara:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:38:36\">00:38:36</a>, a major shift is introduced through Project Solara. Stevie Bathiche and team reveal this as the platform for new \"agent-first\" computing—an ecosystem that connects specialized devices through Azure. Two prototypes, a stationary agent terminal and a portable smart badge, showcase how agents interact contextually. Demonstrations show voice and vision-based real-time intelligence, with implications across industries from healthcare to retail. Qualcomm’s Cristiano Amon joins remotely, discussing silicon architectures adapted for distributed intelligence. Solara signals the evolution from single-device computing to interconnected personal agents functioning seamlessly across environments—a key milestone toward an agentic platform layer spanning cloud, device, and edge.</p>\n\n<p><strong>From Local Agents to Foundry and Frontier AI:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"01:16:00\">01:16:00</a>, the focus returns to Microsoft Foundry, the platform for building, hosting, and governing long-running AI agents. GitHub Copilot evolves into a full application layer, connecting with Foundry and the Rayfin SDK to offer backend services for coding agents. Live demos show creation of persistent, autonomous workflows integrated into enterprise systems. Later, Foundry’s governance layer, Agent 365, ensures containment, security, and compliance across all agentic systems. In <a class=\"medius-aisummary\" data-starttime=\"01:49:00\">01:49:00</a>, Mustafa Suleyman introduces the “Mai” model family—new reasoning, imaging, and voice models powering this ecosystem. The introduction of Frontier Tuning enables developers and enterprises to create bespoke reinforcement learning environments, turning their operational data into proprietary “hill climbing” intelligence tuned to their business objectives.</p>\n\n<p><strong>Scientific Discovery, Quantum Advancements, and Closing Vision:</strong> At <a class=\"medius-aisummary\" data-starttime=\"02:11:09\">02:11:09</a>, Satya presents Microsoft Discovery, a revolutionary platform combining AI agents, HPC, and lab automation to accelerate real-world scientific progress. A live demonstration shows Discovery autonomously hypothesizing, modeling, and validating new proteins for sustainable plastics recycling—linking digital AI reasoning with automated wet labs. This merges physical and computational domains, redefining discovery. The keynote concludes with the unveiling of the Majorana 2 quantum chip at <a class=\"medius-aisummary\" data-starttime=\"02:19:04\">02:19:04</a>, achieving breakthrough reliability and density for the coming quantum era. Nadella closes by reaffirming Microsoft’s north star: using frontier technologies—AI, autonomous agents, and quantum computing—to expand human opportunity responsibly, making this ecosystem a collective engine for progress rather than concentrated power.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9b6fd85d-e64f-46a4-b8b2-23811087192f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9b6fd85d-e64f-46a4-b8b2-23811087192f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9b6fd85d-e64f-46a4-b8b2-23811087192f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9b6fd85d-e64f-46a4-b8b2-23811087192f","speakerIds":["1768245294609build26-1775670526436001p1Pd"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.894Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9b6fd85d-e64f-46a4-b8b2-23811087192f","ogTitle":"Microsoft Build opening keynote","ogDescription":"Satya Nadella and Microsoft leaders share how Microsoft is creating new opportunity for developers across our platforms in this era of AI.\n\nThere is limited seating in the Festival Pavillion for the keynote. Please arrive early to secure a spot. Additional keynote viewing areas will be available in the outdoor theater, BATS Theater (building B), Cowell Theater, and inside Gateway Pavillion, level 2, at all theaters.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion","logicalValue":"Festival Pavilion"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775670526436001p1Pd","RegistrantKey":"1768245294609build26-1775670526436001p1Pd"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"e1f72827-c1f1-44d0-b3b0-16a16aca1688","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778527811637001WS8E","langLocale":"en-US","sessionId":"1778527811637001WS8E","sessionInstanceId":"1778527811637001WS8E","sessionCode":"KEY01-O1","sessionCodeNormalized":"EN-US-KEY01-O1","title":"Overflow Keynote Viewing: BATS Improv","sortTitle":"overflow keynote viewing: bats improv","description":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527811637001WS8E_20260513131810588.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.228Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527811637001WS8E_20260513131810588.avif","ogTitle":"Overflow Keynote Viewing: BATS Improv","ogDescription":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"f739e7f0-34d0-4798-b1d7-19e12680816a","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778527729376001IFme","langLocale":"en-US","sessionId":"1778527729376001IFme","sessionInstanceId":"1778527729376001IFme","sessionCode":"KEY01-O2","sessionCodeNormalized":"EN-US-KEY01-O2","title":"Overflow Keynote Viewing: Cowell Theater","sortTitle":"overflow keynote viewing: cowell theater","description":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527729376001IFme_20260513131801930.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.777Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527729376001IFme_20260513131801930.avif","ogTitle":"Overflow Keynote Viewing: Cowell Theater","ogDescription":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"0e16d0a3-3c61-4d48-b61c-67f34f61a5a7","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778527797402001uj2B","langLocale":"en-US","sessionId":"1778527797402001uj2B","sessionInstanceId":"1778527797402001uj2B","sessionCode":"KEY01-O5","sessionCodeNormalized":"EN-US-KEY01-O5","title":"Overflow Keynote Viewing: Festival Pavilion, Theater A","sortTitle":"overflow keynote viewing: festival pavilion, theater a","description":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527797402001uj2B_20260521160548581.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.235Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527797402001uj2B_20260521160548581.avif","ogTitle":"Overflow Keynote Viewing: Festival Pavilion, Theater A","ogDescription":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n \nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"436cdfb6-2ca3-4fbd-b3a9-9b55eaf354c4","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778527633559001BZRp","langLocale":"en-US","sessionId":"1778527633559001BZRp","sessionInstanceId":"1778527633559001BZRp","sessionCode":"KEY01-O3","sessionCodeNormalized":"EN-US-KEY01-O3","title":"Overflow Keynote Viewing: Gateway Pavilion","sortTitle":"overflow keynote viewing: gateway pavilion","description":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n\nOverflow Keynote Viewing is available in the following spaces in Gateway Pavilion:\n• Gateway Pavilion, Level 1, Cowell Theater\n• Gateway Pavilion, Level 2, Theater B\n• Gateway Pavilion, Level 2, Theater C\n• Gateway Pavilion, Level 2, Theater 1\n• Gateway Pavilion, Level 2, Theater 2\n• Gateway Pavilion, Level 2, Theater 3\n\nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527633559001BZRp_20260513131744288.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.776Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527633559001BZRp_20260513131744288.avif","ogTitle":"Overflow Keynote Viewing: Gateway Pavilion","ogDescription":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n\nOverflow Keynote Viewing is available in the following spaces in Gateway Pavilion:\n• Gateway Pavilion, Level 1, Cowell Theater\n• Gateway Pavilion, Level 2, Theater B\n• Gateway Pavilion, Level 2, Theater C\n• Gateway Pavilion, Level 2, Theater 1\n• Gateway Pavilion, Level 2, Theater 2\n• Gateway Pavilion, Level 2, Theater 3\n\nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion","logicalValue":"Gateway Pavilion"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"39e44792-d729-4386-a976-554a23255a62","startDateTime":"2026-06-02T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778527634258001BXDV","langLocale":"en-US","sessionId":"1778527634258001BXDV","sessionInstanceId":"1778527634258001BXDV","sessionCode":"KEY01-O4","sessionCodeNormalized":"EN-US-KEY01-O4","title":"Overflow Keynote Viewing: Outdoor Theater","sortTitle":"overflow keynote viewing: outdoor theater","description":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n\nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":1,"TimeSlot":"16:30 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:00:00Z","durationInMinutes":150,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527634258001BXDV_20260513131754370.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.777Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778527634258001BXDV_20260513131754370.avif","ogTitle":"Overflow Keynote Viewing: Outdoor Theater","ogDescription":"Due to limited seating in the Festival Pavillion, this location is available for overflow keynote viewing. All overflow locations are first-come, first-served. \n\nLive captions will be available in all keynote overflow viewing locations. American Sign Language (ASL) interpretation will be available in select locations, including Cowell Theater, BATS Theater, and the Outdoor Theater.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Keynote","logicalValue":"Keynote"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Outdoor Theater","logicalValue":"Outdoor Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"}]},{"@search.score":1.0,"documentId":"81ef5694-3fc6-4855-a22f-342d7f0865b7","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Addy Osmani","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780003071986001anuV","langLocale":"en-US","sessionId":"1780003071986001anuV","sessionInstanceId":"1780003071986001anuV","sessionCode":"LIVE150","sessionCodeNormalized":"EN-US-LIVE150","title":"Agent Skills","sortTitle":"agent skills","description":"In this session, Addy joins us to talk about what agent skills actually are, how they encode senior engineering judgment into workflows any AI agent can follow, and how to use them to get consistent, production-quality results out of your coding tools.","aiDescription":"<p><strong>Introduction and Guest Overview:</strong> The video opens at <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a> with a friendly reunion between the host and Addy Osmani, Director at Google Cloud AI and best-selling author. The host reminisces about Addy’s presence in an AI Bloomberg segment and their long history going back to jQuery and web performance discussions. Addy is introduced as an expert in AI developer and user experience, here to talk about his recent work on agent skills, a project that has appeared multiple times on Hacker News. This sets the stage for an in-depth look at his skill-based AI engineering approach and a practical demo.</p>\n\n<p><strong>Concept of Agent Skills and Primer:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:37\">00:02:37</a>, Addy explains the notion of agent skills — standardized modular packages of expertise that expand an AI agent’s capabilities. He relates them to developer \"dot files\" or \"magic tricks,\" reusable personalized tools that others can learn from or copy. Skills, he notes, are reusable across projects and can load dynamically based on the agent’s needs. He introduces a humorous “Clippy” demo — a playful Windows desktop simulation inside a browser — to illustrate how structured skill definitions with a name and description guide the loading process. These standardized skill files allow agents to perform workflows repeatedly across tasks at Google Cloud.</p>\n\n<p><strong>Project Background and Skill Architecture:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:04:14\">00:04:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:50\">00:07:50</a>, Addy discusses his open-source initiative, Agent Skills, inspired by prior frameworks like Gary Tan’s G-Stack. He slowed his open-source activity for some years but decided to release his perspective aligned with the software development life cycle (SDLC). The project organizes skills following stages like define, plan, build, verify, review, and ship. Addy’s approach reflects disciplined agentic engineering, emphasizing quality verification and structured development. Examples include definition skills that clarify ideas, planning and building skills for UX and APIs, and verification skills for automatic browser testing and debugging. Review and shipping stages integrate security, code simplification, CI workflows, and documentation — all embodied as agent-driven skill files. Each skill file includes rationalizations, red flags, and verification criteria to help ensure agents act correctly and efficiently.</p>\n\n<p><strong>Live Demo: Building with Agent Skills:</strong> Between <a class=\"medius-aisummary\" data-starttime=\"00:08:11\">00:08:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>, Addy demonstrates building a “Streak” habit tracker entirely via agent skills using VS Code’s Copilot and Gemini 3.5 Flash. He begins with a raw prompt for a GitHub-inspired tracker and uses the “refine” command to gather clarifying questions, audience definitions, feature constraints, and success criteria. Through iterative refinement, the agent produces validated MVP directions. Addy then moves to “spec,” where a detailed technical specification emerges automatically. Using the “plan” skill, the agent breaks the project into executable tasks with acceptance criteria, mirroring agile development practice. Addy walks through the testing cycles based on TDD principles — red/green test patterns — demonstrating how agents scaffold logic and UI progressively while he explains balancing automation with manual understanding and cognitive load management. The conversation touches productivity philosophies, code review loops, and the trade-offs between speed and discipline in agent-assisted engineering.</p>\n\n<p><strong>Verification, Review, and Security Discussion:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:37:00\">00:37:00</a>, Addy shows the “verify” and “review” phases in action. The agent runs automated browser tests to confirm functionality and UI integrity, then performs multi-pass code reviews covering performance, readability, accessibility, and security. Using skills, Addy encodes best practices to avoid vulnerabilities such as XSS or API leaks. The discussion deepens into reflections on deterministic verification in AI-assisted coding and the industry’s need for stronger security gates. Addy recommends leveraging encoded red flags and deterministic validation to ensure safe deployment. He emphasizes encoding quality standards — accessibility, security, and performance thresholds — directly into skill definitions so agents automatically uphold software reliability.</p>\n\n<p><strong>Q&A and Closing Remarks:</strong> In the final segment at <a class=\"medius-aisummary\" data-starttime=\"00:40:00\">00:40:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:40\">00:45:40</a>, audience questions explore multi-model workflows between Gemini Flash and Pro, using lighter models for planning and robust models for implementation. Addy and the host discuss MCP (Model Context Protocol) integration, authentication, CLI alternatives, and the advantages of progressively-loaded skills for context efficiency and token savings. They conclude that skills offer scalable, modular, and economical AI workflow management. The host thanks Addy for an insightful walkthrough on agent skills and practical AI engineering at scale, wrapping up the session with appreciation and an invitation for attendees to converse further after the talk.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/71402de9-8577-4995-8601-383d2da82704","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/71402de9-8577-4995-8601-383d2da82704","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/71402de9-8577-4995-8601-383d2da82704","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/71402de9-8577-4995-8601-383d2da82704","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1778637888075001yS4N"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/71402de9-8577-4995-8601-383d2da82704","ogTitle":"Agent Skills","ogDescription":"In this session, Addy joins us to talk about what agent skills actually are, how they encode senior engineering judgment into workflows any AI agent can follow, and how to use them to get consistent, production-quality results out of your coding tools.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778637888075001yS4N","RegistrantKey":"1768245294609build26-1778637888075001yS4N"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/AgentSkills-AO","backpackLink":null,"label":"Agent skills","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"9fb3a62c-0bb1-452f-aa95-ce529a0b32fd","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"The Chainsmokers, Charles Lamanna","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779989182067001ripA","langLocale":"en-US","sessionId":"1779989182067001ripA","sessionInstanceId":"1779989182067001ripA","sessionCode":"LIVE110","sessionCodeNormalized":"EN-US-LIVE110","title":"Backing the future of innovation","sortTitle":"backing the future of innovation","description":"Join Charles Lamanna with the founders of Mantis VC—also known as The Chainsmokers—for a candid conversation on what’s shaping the next wave of innovation.\n \nFrom building and backing companies to spotting early signals in AI, the discussion will explore what’s gaining real traction, what’s overhyped, and where builders should focus next. Expect honest perspectives on how technology, culture, and go-to-market distribution come together to turn ideas into lasting products.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session begins with a friendly welcome and introductions around <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>. The moderator expresses excitement to host Drew and Alex, known both as members of The Chainsmokers and as leaders of Mantis VC. They explain how their work now intersects music, technology, and startups. The guests highlight that they have been actively investing for years and are about to launch their fourth venture fund <a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>. Their enthusiasm centers on technology and entrepreneurship, setting the tone for a conversation exploring the evolution of startups and the future driven by AI.</p>\n\n<p><strong>Investment Philosophy and Early-Stage Focus:</strong> In the discussion about what they look for when investing, Drew and Alex emphasize that early-stage investing is about backing “the right people” rather than solely metrics or empirical data, since these young companies lack proven records <a class=\"medius-aisummary\" data-starttime=\"00:01:31\">00:01:31</a>. Their philosophy centers on supporting mission-driven founders who envision where the world is heading, even if their initial predictions are off. They note that great outcomes often result from partnering with resilient, visionary teams. This portion <a class=\"medius-aisummary\" data-starttime=\"00:01:56\">00:01:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:09\">00:02:09</a> captures their conviction that founder quality drives long-term success more than specific ideas at inception, and they describe how the capabilities of small teams are expanding dramatically with the emergence of AI “agents” that amplify efficiency across industries.</p>\n\n<p><strong>Shifting Dynamics with AI and Frontier Technology:</strong> As the conversation deepens around <a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>, they discuss how AI is creating new leverage in businesses, enabling smaller teams to scale faster. The speakers argue that technology can make startups more vertically capable and efficient. Alex challenges the notion of a “one-person billion-dollar company” by underscoring the importance of teamwork and partnerships despite increased automation <a class=\"medius-aisummary\" data-starttime=\"00:03:36\">00:03:36</a>. The dialogue explores how AI lets founders pursue harder, frontier-tech problems—areas once inaccessible due to cost or complexity. This evolution excites them as they see elite talent increasingly moving into frontier AI spaces, such as autonomous agents and complex system architectures that require continuous rethinking of governance, security, and development practices <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:32\">00:07:32</a>.</p>\n\n<p><strong>Practical Use of AI Tools and Workflow Integration:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:08:04\">00:08:04</a>, the conversation turns personal with questions about how they use AI tools. Drew mentions leveraging AI heavily for summarizing information and research, improving his productivity and comprehension of complex topics. In contrast, Alex notes they do not currently use AI in music production since their creative process relies on organic collaboration. However, within their venture fund, they have applied AI through systems like Palantir to improve due diligence and operational intelligence <a class=\"medius-aisummary\" data-starttime=\"00:09:12\">00:09:12</a>. The group discusses integrating AI into existing workflows—rather than forcing new tool adoption—and applauds the growing movement to bring AI into spaces people already work. This section highlights their belief that AI should amplify human talent, not replace it.</p>\n\n<p><strong>Investment Thesis on AI Layers and Vertical Applications:</strong> Starting near <a class=\"medius-aisummary\" data-starttime=\"00:10:26\">00:10:26</a>, the discussion dives into where value will accrue within the tech stack. Drew and Alex share that while they have personal investments in major model providers like Anthropic and OpenAI, they expect those layers to become commoditized. Their fund focuses on highly verticalized AI applications in regulated domains like finance, legal, enterprise compliance, and healthcare <a class=\"medius-aisummary\" data-starttime=\"00:11:33\">00:11:33</a>. They describe investments in companies such as Monaco AI and Kepler that build defensible products through regulatory expertise and proprietary data. The conversation addresses the balance between foundational model providers and application layer innovators, predicting strong opportunities higher up the stack where specialized knowledge and data create enduring moats <a class=\"medius-aisummary\" data-starttime=\"00:13:24\">00:13:24</a>.</p>\n\n<p><strong>Culture, Creativity, and Long-Term Resilience:</strong> In the closing segment <a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>, Drew and Alex draw parallels between their artistic journey and startup building. They reflect on their early song releases, including one that succeeded commercially but didn’t represent their authentic values. They relate this to founders who pursue opportunities that don’t fit long-term vision and emphasize maintaining clarity of purpose. They advise startup teams to focus on intentional creation, learn quickly from failure, and cultivate a loyal “cult following” around meaningful work. Authenticity, adaptability, and obsession with one’s mission are portrayed as essential traits for survival through volatile technological change. The conversation closes with optimistic commentary about building for the future of AI—investing in founders and ideas anticipating what’s next rather than what’s now—illustrated by their investment in Factory AI <a class=\"medius-aisummary\" data-starttime=\"00:19:22\">00:19:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:21\">00:20:21</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"22:15 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:35:00Z","durationInMinutes":20,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b149b982-e4e6-475e-8a2d-8c9391b83ab7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b149b982-e4e6-475e-8a2d-8c9391b83ab7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b149b982-e4e6-475e-8a2d-8c9391b83ab7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b149b982-e4e6-475e-8a2d-8c9391b83ab7","speakerIds":["1768245294609build26-1780418227989001gBHP","1768245294609build26-1750281479896001GXs4"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.24Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b149b982-e4e6-475e-8a2d-8c9391b83ab7","ogTitle":"Backing the future of innovation","ogDescription":"Join Charles Lamanna with the founders of Mantis VC—also known as The Chainsmokers—for a candid conversation on what’s shaping the next wave of innovation.\n \nFrom building and backing companies to spotting early signals in AI, the discussion will explore what’s gaining real traction, what’s overhyped, and where builders should focus next. Expect honest perspectives on how technology, culture, and go-to-market distribution come together to turn ideas into lasting products.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1750281479896001GXs4","RegistrantKey":"1768245294609build26-1750281479896001GXs4"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780418227989001gBHP","RegistrantKey":"1768245294609build26-1780418227989001gBHP"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"a00f73df-ef23-42ca-8f66-cadf2c58966a","startDateTime":"2026-06-02T23:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Monica Cisneros, Scott Hanselman","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779992320790001GAgx","langLocale":"en-US","sessionId":"1779992320790001GAgx","sessionInstanceId":"1779992320790001GAgx","sessionCode":"LIVE144","sessionCodeNormalized":"EN-US-LIVE144","title":"Behind the Keynote: How Windows Made OpenClaw Work in the Keynote Demo","sortTitle":"behind the keynote: how windows made openclaw work in the keynote demo","description":"A candid 15 minute interview with Monica Cisneros and Scott Hanselman on what it took to get the OpenClaw keynote demo ready. Hear how teams across Windows aligned on platform work, runtime integration, and demo engineering to deliver a reliable live experience.","aiDescription":"<p><strong>Introduction and Collaboration Story:</strong> The video begins with Monica Cisneros introducing herself as the Senior Product Marketing Manager for Windows, joined by Scott Hanselman (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>). Monica recalls seeing Scott at the keynote, where he presented a conversation with Peter and asks him to describe how that collaboration evolved. Scott explains that open-source success rarely happens overnight — Peter had already created hundreds of innovative utilities and was recognized for his contributions within the open-source space (<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>). Seeing Peter’s project, Scott wanted to build a Windows counterpart that mirrored the Mac app, showing that collaboration can often start with a simple shared idea (<a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:22\">00:01:22</a>).</p>\n\n<p><strong>The Essence of Open-Source Community:</strong> Monica asks what makes the open-source community special (<a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:42\">00:02:42</a>). Scott explains that trust and familiarity are built through repeated encounters at meetups or online, noting that seeing people consistently contributes to feeling safe within the community (<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>). He describes the environment as collaborative and improvisational — similar to improv comedy where participants respond with “Yes, and…” instead of “No” (<a class=\"medius-aisummary\" data-starttime=\"00:03:51\">00:03:51</a>). Scott emphasizes that open-source thrives on that positivity, where one person’s idea sparks contributions from others, resulting in cooperative building across different skills and energy levels (<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:09\">00:04:09</a>).</p>\n\n<p><strong>Working Philosophy and Agency:</strong> Transitioning from community to workflow, Scott discusses the concept of “agency” in open-source production (<a class=\"medius-aisummary\" data-starttime=\"00:04:35\">00:04:35</a>). He recounts Vincent Koch’s statement that he likes “high-agency people” who take initiative to fix issues rather than waiting for approval (<a class=\"medius-aisummary\" data-starttime=\"00:04:50\">00:04:50</a>). This mindset, alongside robust testing and quality assurance, creates a safety net for experimentation within the Open Claw Foundation project. Scott encourages engineers to be bold — to take control, make changes confidently, and rely on automated testing to ensure stability, a vital attitude for scalable AI-augmented software projects (<a class=\"medius-aisummary\" data-starttime=\"00:05:12\">00:05:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:44\">00:05:44</a>).</p>\n\n<p><strong>Building the Windows Companion App:</strong> When asked about engineering lessons learned from the Windows companion app effort, Scott explains that the project aimed for parity with the existing Mac application but soon explored what Windows could do differently (<a class=\"medius-aisummary\" data-starttime=\"00:06:25\">00:06:25</a>). The team examined app packaging, permissions, and sandboxing unique to Windows, approaching the process as “coopetition” — improving all platforms through mutual advancement (<a class=\"medius-aisummary\" data-starttime=\"00:06:51\">00:06:51</a>). He describes how Microsoft is adopting a faster iterative style while keeping stability for its massive user base, balancing innovation and scale to keep all users satisfied (<a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:52\">00:07:52</a>).</p>\n\n<p><strong>Containers and Containment Concepts:</strong> The conversation moves to Scott’s keynote remarks about containers and containment (<a class=\"medius-aisummary\" data-starttime=\"00:07:53\">00:07:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:02\">00:08:02</a>). He explains that containers enable isolated environments using technologies like Docker, but containment is a broader concept — defining boundaries around what software can access (<a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>). With WSL-based container runtimes, Windows now supports these environments natively (<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>). Scott gives practical examples — isolating access between work applications — to illustrate how containment enforces responsible permissions, preventing unintended cross-app data access (<a class=\"medius-aisummary\" data-starttime=\"00:09:20\">00:09:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:40\">00:09:40</a>).</p>\n\n<p><strong>Modern Developer Workflows and the Future of Agents:</strong> Scott shares insights into modern workflows like Git work trees that allow multiple parallel branches and agents to operate simultaneously (<a class=\"medius-aisummary\" data-starttime=\"00:11:05\">00:11:05</a>). This parallelism, coupled with tools such as the GitHub Copilot app integrated with Windows dev drives, enhances productivity and development speed (<a class=\"medius-aisummary\" data-starttime=\"00:11:24\">00:11:24</a>). Discussing the future, Scott predicts dynamic policy engines that tailor containment and access on a per-tool-call basis (<a class=\"medius-aisummary\" data-starttime=\"00:12:17\">00:12:17</a>). The goal is frictionless trust and control — agents that act proactively within clear boundaries invisible to users but governed by IT, ensuring security and efficiency (<a class=\"medius-aisummary\" data-starttime=\"00:12:53\">00:12:53</a>). The video concludes with Monica thanking Scott for his intentional and impactful approach to advancing agent technology on Windows (<a class=\"medius-aisummary\" data-starttime=\"00:13:10\">00:13:10</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"23:40 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/710e231b-0fbe-47fb-8681-ad4344165d34","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/710e231b-0fbe-47fb-8681-ad4344165d34","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/710e231b-0fbe-47fb-8681-ad4344165d34","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/710e231b-0fbe-47fb-8681-ad4344165d34","speakerIds":["1768245294609build26-1769028763361001wc2X","1768245294609build26-1769008130185001WAvh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.459Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/710e231b-0fbe-47fb-8681-ad4344165d34","ogTitle":"Behind the Keynote: How Windows Made OpenClaw Work in the Keynote Demo","ogDescription":"A candid 15 minute interview with Monica Cisneros and Scott Hanselman on what it took to get the OpenClaw keynote demo ready. Hear how teams across Windows aligned on platform work, runtime integration, and demo engineering to deliver a reliable live experience.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769028763361001wc2X","RegistrantKey":"1768245294609build26-1769028763361001wc2X"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"f67e4fb3-902f-405f-94c1-f7aa1b9f052a","startDateTime":"2026-06-03T21:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vivek Bhadauria, Nitya Narasimhan, Filisha Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780010031014001eHNr","langLocale":"en-US","sessionId":"1780010031014001eHNr","sessionInstanceId":"1780010031014001eHNr","sessionCode":"LIVE159","sessionCodeNormalized":"EN-US-LIVE159","title":"Behind the Scenes: Accelerating the AI Agent DevOps Lifecycle with End-to-End Observability","sortTitle":"behind the scenes: accelerating the ai agent devops lifecycle with end-to-end observability","description":"In this interview we unpack what it really took to deliver our end‑to‑end “observe → evaluate → optimize” flow covering the end-to-end Agent DevOps lifecycle from inner‑loop offline signals to continuous improvement in production.  We’ll share hard-earned lessons on (1) simplifying the getting started experience with out of the box observability powered by context-specific eval rubrics, (2) streamlining the developer experience with guided skill-based flows and (3) leveraging the complete set of inner and outer loop signals for continuous improvement.","aiDescription":"<p><strong>Introductions and Session Overview:</strong> The session opens with greetings and introductions as the host, Nithya, welcomes the audience and her co-presenters Felicia and Vivek <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>. Both Felicia and Vivek are members of the Foundry Observability team, and the conversation begins by setting the stage for a discussion on agent optimization—an advanced topic within AI development. Nithya expresses excitement about diving into why agent optimization is challenging and how the Foundry platform helps make the process more efficient for developers managing complex, stateful, and adaptive software agents <a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>.</p>\n\n<p><strong>Challenges in Agent Optimization:</strong> Vivek explains that agents differ from traditional software because they are stateful, long-horizon systems capable of planning, adapting, and interacting dynamically with their environment <a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:49\">00:00:49</a>. This flexibility makes agents powerful but also very difficult to debug, test, or reproduce reliably. Foundry’s Agent platform provides an integrated environment where users can build, run, trace, evaluate, and optimize their agents within a single unified system <a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:27\">00:01:27</a>. Nithya and Felicia emphasize how essential observability is across an agent’s development pipeline—from planning through deployment—to ensure developers can understand why models behave as they do and can iteratively improve performance.</p>\n\n<p><strong>Evaluators and Evaluation Rubrics:</strong> The discussion transitions into how Foundry introduces evaluators, tools designed to enable consistent testing of non-deterministic systems <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>. Felicia explains that evaluators act as a new form of test-driven development for AI—allowing agents to be assessed using automated, data-driven criteria. These evaluators can be applied both in the development phase (inner loop) and in production environments (outer loop), offering developers continuous feedback on performance. The group then explores “Eval Rubrics” <a class=\"medius-aisummary\" data-starttime=\"00:02:52\">00:02:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:25\">00:03:25</a>, a multidimensional assessment framework built by Foundry. Eval Rubrics enable weighted and domain-specific evaluations that help quantify an agent’s reliability, adaptability, and decision accuracy, thus serving as powerful guardrails for complex behaviors.</p>\n\n<p><strong>Continuous Optimization and Production Readiness:</strong> The conversation turns to how agents evolve after entering production <a class=\"medius-aisummary\" data-starttime=\"00:03:45\">00:03:45</a>. Vivek describes how developers can apply continuous optimization using Foundry’s “Agent Optimizer” tool <a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>, which ensures that agents keep improving based on live data and real-world interaction traces. The process involves maintaining high-quality “golden datasets” and consistent evaluators to detect regressions before updates are deployed. Over time, as agents run more sessions and collect more operational data, they become smarter and more robust. The panel notes that “Agent Optimizer” is currently in gated preview and will soon be available in public preview <a class=\"medius-aisummary\" data-starttime=\"00:05:49\">00:05:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:57\">00:05:57</a>, inviting developers to try it and provide feedback.</p>\n\n<p><strong>Advanced Evaluation and Trace Tools:</strong> Nithya highlights additional Foundry capabilities like Foundry Skills, support for multi-turn conversational evaluation, and user simulation tools that allow developers to diagnose failed traces or unexpected outputs quickly <a class=\"medius-aisummary\" data-starttime=\"00:06:10\">00:06:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:39\">00:06:39</a>. Features like Trace Replace simplify navigating complex interaction logs and reproducing issues efficiently. Collectively, these advancements demonstrate Foundry’s end-to-end observability model—spanning from agent creation to adaptive optimization—strengthening confidence in large, intelligent systems operating at scale.</p>\n\n<p><strong>Reflections and Closing Remarks:</strong> As time runs out <a class=\"medius-aisummary\" data-starttime=\"00:07:08\">00:07:08</a>, Nithya invites the panelists to share their personal takeaways from the Build event. Vivek reflects on how energizing it is to engage directly with developers and see how Foundry’s tools address real-world challenges <a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>, while Felicia expresses appreciation for the in-person collaboration and the vibrant atmosphere. The session closes with a reminder for viewers to check out breakout session “BRK 252” to see a live demo of Agent Optimizer in action <a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:36\">00:08:36</a>, wrapping up a rich discussion on how Foundry is helping developers continuously build, monitor, and evolve their AI agents.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:10 - 21:20","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:20:00Z","durationInMinutes":10,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/47bd4df9-ea64-49d9-bd10-d4e2f5baefcc","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/47bd4df9-ea64-49d9-bd10-d4e2f5baefcc","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/47bd4df9-ea64-49d9-bd10-d4e2f5baefcc","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/47bd4df9-ea64-49d9-bd10-d4e2f5baefcc","speakerIds":["1768245294609build26-1778597318722001S00J","1768245294609build26-1756141269010001LleM","1768245294609build26-1774284995366001pNmt"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/47bd4df9-ea64-49d9-bd10-d4e2f5baefcc","ogTitle":"Behind the Scenes: Accelerating the AI Agent DevOps Lifecycle with End-to-End Observability","ogDescription":"In this interview we unpack what it really took to deliver our end‑to‑end “observe → evaluate → optimize” flow covering the end-to-end Agent DevOps lifecycle from inner‑loop offline signals to continuous improvement in production.  We’ll share hard-earned lessons on (1) simplifying the getting started experience with out of the box observability powered by context-specific eval rubrics, (2) streamlining the developer experience with guided skill-based flows and (3) leveraging the complete set of inner and outer loop signals for continuous improvement.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778597318722001S00J","RegistrantKey":"1768245294609build26-1778597318722001S00J"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995366001pNmt","RegistrantKey":"1768245294609build26-1774284995366001pNmt"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d716c6e3-7780-4e19-90bd-3540e56132e7","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Satya Nadella, swyx (Shawn Wang), Elad Gil, Sarah Guo","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780092525426001glwE","langLocale":"en-US","sessionId":"1780092525426001glwE","sessionInstanceId":"1780092525426001glwE","sessionCode":"LIVE100","sessionCodeNormalized":"EN-US-LIVE100","title":"Build 2026 podcast with Satya Nadella, Sarah Guo and Elad Gil from No Priors, and Swyx from Latent Space","sortTitle":"build 2026 podcast with satya nadella, sarah guo and elad gil from no priors, and swyx from latent space","description":"Satya Nadella sits down for an interview after the Build 2026 keynote with Sarah Guo and Elad Gil from No Priors, and Swyx from Latent Space to discuss Microsoft news, covering AI and coding innovation, Windows hardware and major updates to the Windows developer experience, agentic AI breakthroughs, and more.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b78416d2-cb5f-4e7c-8b39-f3c7404caf3c","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b78416d2-cb5f-4e7c-8b39-f3c7404caf3c","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b78416d2-cb5f-4e7c-8b39-f3c7404caf3c","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b78416d2-cb5f-4e7c-8b39-f3c7404caf3c","speakerIds":["1768245294609build26-1775670526436001p1Pd","1768245294609build26-1775516734838001pZmI","1768245294609build26-1779800220553001ArTL","1768245294609build26-1779817062828001z6pY"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b78416d2-cb5f-4e7c-8b39-f3c7404caf3c","ogTitle":"Build 2026 podcast with Satya Nadella, Sarah Guo and Elad Gil from No Priors, and Swyx from Latent Space","ogDescription":"Satya Nadella sits down for an interview after the Build 2026 keynote with Sarah Guo and Elad Gil from No Priors, and Swyx from Latent Space to discuss Microsoft news, covering AI and coding innovation, Windows hardware and major updates to the Windows developer experience, agentic AI breakthroughs, and more.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775670526436001p1Pd","RegistrantKey":"1768245294609build26-1775670526436001p1Pd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775516734838001pZmI","RegistrantKey":"1768245294609build26-1775516734838001pZmI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779800220553001ArTL","RegistrantKey":"1768245294609build26-1779800220553001ArTL"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779817062828001z6pY","RegistrantKey":"1768245294609build26-1779817062828001z6pY"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"ea15754a-9a28-466d-97c3-36263b20b659","startDateTime":"2026-06-02T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nikola Metulev, Beth Pan, Aditya Ramnathkar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779988033906001rQXU","langLocale":"en-US","sessionId":"1779988033906001rQXU","sessionInstanceId":"1779988033906001rQXU","sessionCode":"LIVE172","sessionCodeNormalized":"EN-US-LIVE172","title":"Build and ship faster with a developer-optimized Windows experience","sortTitle":"build and ship faster with a developer-optimized windows experience","description":"Take a closer look at a developer-optimized Windows experience built to help you move faster. You’ll see how streamlined workflows across WSL, Terminal, WinGet, and your favorite tools reduce friction, leverage agents, build repeatable scenarios, and scale AI projects with confidence.","aiDescription":"<p><strong>Introduction and Windows Developer First Vision:</strong> The session opens with Aditya Ramnathkar, joined by software engineers Beth Pan and Nicola Matulov, introducing themselves as members of the Windows Platform team at Microsoft Build <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>. Aditya sets the stage by explaining that the focus of Windows 11 is to become the best operating system for developers — distraction-free, streamlined, and developer-first. He highlights a series of new improvements launched during Build: Core Utilities support, native Linux containers via WSL, an effortless one-command setup for developer environments, agent-augmented coding capabilities, an intelligent terminal, and next-generation AI-ready hardware for sustainable workloads <a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:49\">00:00:49</a>. The goal is to help developers build and ship faster on Windows.</p>\n\n<p><strong>Streamlined Developer Setup with Winget Configurations:</strong> Beth continues the discussion by addressing a common pain point reported by developers: the difficult onboarding and frequent setup overhead when switching or refreshing machines <a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:22\">00:01:22</a>. To solve this, Windows now provides an automated setup method using Winget configuration files that can install preferred tools like Python, Node.js, PowerToys, and more in a distraction-free environment <a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a>. All configurations are hosted on a public GitHub repository for flexibility and easy customization. Developers can fork and modify these configurations as needed. Beth clarifies that the process eliminates repetitive steps such as enabling developer mode or manually toggling dark mode and visibility options in File Explorer. Winget configuration can instantly apply all preferences with a single command, offering ready-to-code systems out of the box on new devices or existing ones <a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:44\">00:03:44</a>.</p>\n\n<p><strong>Core Utilities and Native Linux Containers with WSL:</strong> The discussion transitions to the integration of native UNIX/Linux tools within the Windows Terminal <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>. Developers can now use familiar commands like `grep` and `netstat` directly without translation into PowerShell syntax. Scripts imported from UNIX or Mac systems work seamlessly, reducing friction when collaborating across environments. Following this, Nicola introduces “WSL Containers” — a new capability that enables containerized development natively on Windows via the WSLC binary <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:39\">00:06:39</a>. With WSLC, users can create, run, and manage Linux containers directly from Windows without additional tools, supporting local dev workflows, AI/ML workloads, and enterprise testing. The update allows developers to run mixed environments — combining native Windows applications with Linux code – and even includes enterprise management controls for visibility and governance <a class=\"medius-aisummary\" data-starttime=\"00:07:15\">00:07:15</a>.</p>\n\n<p><strong>Agent-Augmented Development and WinApp Tools:</strong> Nicola then presents updates around AI-assisted development for Windows applications <a class=\"medius-aisummary\" data-starttime=\"00:07:39\">00:07:39</a>. Using GitHub Copilot and Windows development plugins, developers can leverage specialized agents and skills for tasks such as packaging, design, and quality checks through Win UI-specific capabilities <a class=\"medius-aisummary\" data-starttime=\"00:08:14\">00:08:14</a>. A new CLI tool named WinApp enables direct app packaging and identity assignment without requiring Visual Studio. This supports development across languages like Rust or .NET while providing access to Windows APIs for notifications and AI integration <a class=\"medius-aisummary\" data-starttime=\"00:09:20\">00:09:20</a>. Moreover, WinApp connects seamlessly with terminal workflows, letting developers inspect running applications, debug UI elements, and trigger visual changes or screenshots—all AI assisted. The ecosystem also introduces template generators for Win UI projects, empowering both human developers and coding agents to quickly generate and debug full Windows applications within one environment <a class=\"medius-aisummary\" data-starttime=\"00:10:31\">00:10:31</a>.</p>\n\n<p><strong>Intelligent Terminal with Built-In Agents:</strong> Building upon AI integration, Nicola highlights the new experimental Intelligent Terminal—a modified Windows Terminal featuring built-in agent capabilities and Copilot integration <a class=\"medius-aisummary\" data-starttime=\"00:12:13\">00:12:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:18\">00:12:18</a>. The terminal can automatically detect and fix errors, suggest corrections for mistyped commands, and even execute automated recovery steps when permitted <a class=\"medius-aisummary\" data-starttime=\"00:12:43\">00:12:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:53\">00:12:53</a>. It supports multiple AI models beyond Copilot and retains full transcription of command history for context-aware assistance. The Intelligent Terminal is optional, permission-based, and installable alongside the standard terminal. This creates a cohesive, productivity-focused development experience where everyday tasks like Git fix-ups, branching issues, and reset commands can be resolved instantly through natural interaction <a class=\"medius-aisummary\" data-starttime=\"00:13:04\">00:13:04</a>.</p>\n\n<p><strong>AI Hardware and Closing Reflections:</strong> The presentation closes with Aditya introducing the new Surface RTX Spark Dev Box <a class=\"medius-aisummary\" data-starttime=\"00:14:22\">00:14:22</a>, a purpose-built developer machine with NVIDIA RTX Spark GPU configured for local AI computation featuring up to 1 petaflop of performance and 128 GB of unified memory. This hardware arrives pre-optimized with all new developer tools and software configurations enabled. The hosts conclude by emphasizing the importance of continuous feedback through the \"Voice of Developer\" initiative <a class=\"medius-aisummary\" data-starttime=\"00:15:59\">00:15:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:14\">00:16:14</a>. They encourage developers to explore these innovations, integrate them into workflows, and share insights for future improvements. The session ends on an optimistic note, reaffirming Microsoft's commitment to building a modern, AI-enabled, and developer-centric Windows ecosystem.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:45 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8eb66ee9-b869-4e25-a1e3-d19c18f30d2b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8eb66ee9-b869-4e25-a1e3-d19c18f30d2b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8eb66ee9-b869-4e25-a1e3-d19c18f30d2b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8eb66ee9-b869-4e25-a1e3-d19c18f30d2b","speakerIds":["1768245294609build26-1774886013101001R9mk","1768245294609build26-1775078130718001cIWp","1768245294609build26-1768432088934001dCEQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.24Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8eb66ee9-b869-4e25-a1e3-d19c18f30d2b","ogTitle":"Build and ship faster with a developer-optimized Windows experience","ogDescription":"Take a closer look at a developer-optimized Windows experience built to help you move faster. You’ll see how streamlined workflows across WSL, Terminal, WinGet, and your favorite tools reduce friction, leverage agents, build repeatable scenarios, and scale AI projects with confidence.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774886013101001R9mk","RegistrantKey":"1768245294609build26-1774886013101001R9mk"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775078130718001cIWp","RegistrantKey":"1768245294609build26-1775078130718001cIWp"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768432088934001dCEQ","RegistrantKey":"1768245294609build26-1768432088934001dCEQ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e97f3bda-40ec-48d4-a16b-26c426009e46","startDateTime":"2026-06-03T20:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Tyler Leonhardt","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780009018356001vbEN","langLocale":"en-US","sessionId":"1780009018356001vbEN","sessionInstanceId":"1780009018356001vbEN","sessionCode":"LIVE168","sessionCodeNormalized":"EN-US-LIVE168","title":"Claude Is in Copilot. Here's What That Actually Means","sortTitle":"claude is in copilot. here's what that actually means","description":"Claude runs as a coding agent inside GitHub Copilot in VS Code. But what does that actually look like at the code level? How is context assembled? What tools does Claude have access to? What happens when you pick Claude in the model picker versus letting Copilot run? Tyler Leonhardt goes inside the integration so you know exactly what you are working with.","aiDescription":"<p><strong>Opening and Introduction:</strong> The session begins with a lighthearted welcome as the hosts re-engage the audience at <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>. They check in energetically with attendees, ensuring everyone remains alert and enthusiastic despite the late afternoon timing. After setting an upbeat tone, the host introduces Tyler from the Visual Studio Code (VS Code) team. The conversation quickly pivots toward the main topic — exploring new features in VS Code that incorporate Claude agents, signaling a shift from the more familiar Copilot focus at <a class=\"medius-aisummary\" data-starttime=\"00:00:30\">00:00:30</a>.</p>\n\n<p><strong>Claude Agent Integration in VS Code:</strong> Tyler explains that the Claude Agent integration is based on Anthropic’s Agent SDK, which works similarly to the Copilot SDK but allows users flexibility with their existing Copilot subscriptions (<a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a>). This integration lets developers choose between multiple AI “harnesses” — Claude, Copilot CLI, local configurations, and Codex. The design offers freedom without requiring different billing or account setups. Tyler highlights that enterprises constantly test different AI harnesses, so this compatibility reduces friction as technologies evolve (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>).</p>\n\n<p><strong>Conversation on Real-World Use and Demonstration Setup:</strong> Adding personality to the discussion, Tyler reveals he organizes local improv events and demonstrates using Copilot and Claude together to generate a website for an event called \"Compliment Fest\" (<a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>). He shows how Claude Agent in VS Code automates web creation tasks using visual and design guidance from a flyer PDF. Confetti animations and fun visuals are included, adding humor and human touch to the demo. This practical showcase illustrates the fluid integration between Claude and Copilot while tying back to creative real-world use cases that go beyond code writing.</p>\n\n<p><strong>Exploring Agent Architecture and Protocols:</strong> The talk becomes more technical as Tyler introduces the “agent host protocol” for managing connections between AI agents and user environments (<a class=\"medius-aisummary\" data-starttime=\"00:06:23\">00:06:23</a>). He explains that this protocol enables remote communication so that agents can run on a host machine while users interact from other clients — including web browsers and mobile devices. Demonstrations show VS Code’s ability to connect locally and remotely through VS Code.dev, highlighting secure access via authentication. This flexibility supports seamless transitions from desktop to browser work environments and provides a vision of decentralized AI development workflows (<a class=\"medius-aisummary\" data-starttime=\"00:07:11\">00:07:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>).</p>\n\n<p><strong>Using Claude in the Cloud:</strong> Tyler continues by showing how Claude can run as a cloud-based agent within VS Code’s settings (<a class=\"medius-aisummary\" data-starttime=\"00:09:24\">00:09:24</a>). Users can toggle Claude or Codex coding agents for cloud execution — nicknamed humorously “Claude in the Cloud.” The demonstration sends test prompts to the agent, showing it performing lightweight actions like code suggestions and file edits directly in the cloud. The hosts joke that high-end computers are handling trivial requests, adding levity while underscoring how AI processes tasks seamlessly across environments. Tyler emphasizes that all customization options available in Claude Code also function within this setup (<a class=\"medius-aisummary\" data-starttime=\"00:10:09\">00:10:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:55\">00:10:55</a>).</p>\n\n<p><strong>Wrap-Up and Acknowledgments:</strong> As the session winds down, the hosts summarize the main takeaways — flexibility between Copilot and Claude agents, integration with existing subscriptions, and the benefits of remote accessibility (<a class=\"medius-aisummary\" data-starttime=\"00:11:04\">00:11:04</a>). Tyler reiterates that developers gain “freedom of choice” across AI tools, allowing continuous customization in VS Code without new costs or barriers. The audience applauds, and the hosts thank Tyler warmly for the insightful demo and engaging delivery. They close playfully by hinting at more stories behind \"Compliment Fest\" before transitioning to the next session at <a class=\"medius-aisummary\" data-starttime=\"00:12:08\">00:12:08</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"20:10 - 20:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/13f19aee-b872-4a65-8156-2ae9842c4178","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/13f19aee-b872-4a65-8156-2ae9842c4178","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/13f19aee-b872-4a65-8156-2ae9842c4178","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/13f19aee-b872-4a65-8156-2ae9842c4178","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1774285000394001pcWF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/13f19aee-b872-4a65-8156-2ae9842c4178","ogTitle":"Claude Is in Copilot. Here's What That Actually Means","ogDescription":"Claude runs as a coding agent inside GitHub Copilot in VS Code. But what does that actually look like at the code level? How is context assembled? What tools does Claude have access to? What happens when you pick Claude in the model picker versus letting Copilot run? Tyler Leonhardt goes inside the integration so you know exactly what you are working with.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000394001pcWF","RegistrantKey":"1768245294609build26-1774285000394001pcWF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/Claude","backpackLink":null,"label":"Use Claude as your coding agent inside GitHub Copilot in VS Code","category":"Web link"},{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Build your VS Code and Copilot skills with free, hands-on courses","category":"Web link"},{"link":"https://aka.ms/VSCode/DBview","backpackLink":null,"label":"Inspect every prompt, tool call, and model decision in real time","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"54f9dc77-83c6-4253-9e7a-00e8cec769fc","startDateTime":"2026-06-02T23:55:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Michael Chiang, John Maeda, Parth Sareen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779992625684001GKLp","langLocale":"en-US","sessionId":"1779992625684001GKLp","sessionInstanceId":"1779992625684001GKLp","sessionCode":"LIVE145","sessionCodeNormalized":"EN-US-LIVE145","title":"Coding and Personal Agents with Ollama","sortTitle":"coding and personal agents with ollama","description":"A live walkthrough of Ollama launch, showing how developers connect local and cloud open models to real-world applications in a single step. See Copilot CLI and OpenClaw set up live, switch between local and cloud models and connect the agent to web search and vision.","aiDescription":"<p><strong>Introduction and Overview:</strong>  \nThe conversation begins with enthusiastic introductions from the Microsoft Build stage, featuring Michael Chang and engineer Parth Serene from the O Llama team <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>. The hosts joke about the audience’s love for the O Llama mascot before shifting to an explanation of O Llama’s purpose. Michael shares that O Llama provides developers the easiest way to run open models either locally or through the cloud <a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>. It caters to users who prefer full local privacy or those needing scalable compute in a cloud-based setup. This establishes the dual flexibility that defines O Llama’s mission in the open AI model ecosystem.</p>\n\n<p><strong>Local Model Agents and Developer Adoption:</strong>  \nThe discussion moves to the evolving landscape of local AI models and how developers have embraced them <a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:32\">00:01:32</a>. Michael notes that coding use cases exploded in 2023, and in the following year, powerful open models and agents became the norm. Parth explains how he uses O Llama agents for personal analyses, like reviewing his own bank statements—highlighting local models’ privacy advantages <a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:26\">00:02:26</a>. The guests emphasize that developers appreciate how these agents can perform sensitive, private tasks locally and still deliver strong coding performance. Michael recalls the moment O Llama became an overnight sensation after launching publicly on GitHub, quickly becoming one of the fastest growing open-source AI projects <a class=\"medius-aisummary\" data-starttime=\"00:02:54\">00:02:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:15\">00:03:15</a>.</p>\n\n<p><strong>Scaling from Local to Cloud:</strong>  \nParth and Michael delve into how developers can seamlessly scale local workloads to the cloud <a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:27\">00:04:27</a>. They describe scenarios where developers start with local models and later move to cloud infrastructure when handling extreme workloads or complex code tasks. O Llama’s design ensures the transition feels nearly identical to the user, maintaining consistency in experience. The team compares the flexibility to “reaching for the cloud,” humorously invoking Toy Story imagery <a class=\"medius-aisummary\" data-starttime=\"00:04:28\">00:04:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:33\">00:04:33</a>. They demonstrate this hybrid model with an example using the Copilot CLI, showcasing how O Llama empowers developers to handle workflow, debugging, and GitHub issue management directly from their terminals without leaving their core environment <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:33\">00:06:33</a>.</p>\n\n<p><strong>Developer Integrations and Real-World Uses:</strong>  \nAfter the demo, the discussion highlights O Llama’s vast integration ecosystem, supporting VS Code, Hermes, and OpenClaw among others <a class=\"medius-aisummary\" data-starttime=\"00:06:33\">00:06:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>. The hosts liken O Llama’s hybrid operation to a hybrid car—switching smoothly between local and cloud resources. Michael and Parth elaborate on the hardest parts of scaling AI apps and how developers commonly start locally before turning to O Llama’s cloud options <a class=\"medius-aisummary\" data-starttime=\"00:07:02\">00:07:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>. They express enthusiasm about local agents running personal tasks, managing emails, calendars, and daily planning—showing tangible impact on users’ lives <a class=\"medius-aisummary\" data-starttime=\"00:07:32\">00:07:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:02\">00:08:02</a>. Parth adds that workloads with private data benefit most from local execution, while heavier coding workloads thrive on cloud models for efficiency and faster execution <a class=\"medius-aisummary\" data-starttime=\"00:08:09\">00:08:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>.</p>\n\n<p><strong>Open Source and Future Outlook:</strong>  \nThe conversation transitions toward how local models fit into the open-source movement <a class=\"medius-aisummary\" data-starttime=\"00:09:05\">00:09:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:50\">00:09:50</a>. Parth explains that the evolution of open models is guided by user trends such as personal agents and privacy-guarded applications. The team previews an upcoming breakout session showcasing live demos and more agentic task examples <a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>. Michael predicts the AI world will continue blending local and cloud models—similar to how personal computers coexist with powerful cloud services—to balance privacy and robust computation <a class=\"medius-aisummary\" data-starttime=\"00:10:21\">00:10:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:34\">00:10:34</a>. His advice to new developers is simple: choose a model, start experimenting, and explore open tools to quickly familiarize themselves with this hybrid AI approach <a class=\"medius-aisummary\" data-starttime=\"00:10:53\">00:10:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:16\">00:11:16</a>.</p>\n\n<p><strong>Design Philosophy and Closing Remarks:</strong>  \nIn the final moments, attention turns to O Llama’s design aesthetic and origin story. Michael shares that the product was intentionally built to be visually minimal—monochromatic and unobtrusive—emphasizing simplicity and developer empowerment <a class=\"medius-aisummary\" data-starttime=\"00:11:25\">00:11:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:56\">00:11:56</a>. Parth recounts how his initial admiration for O Llama led him to join the team full-time after building a startup that relied on the tool <a class=\"medius-aisummary\" data-starttime=\"00:12:02\">00:12:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:23\">00:12:23</a>. The segment closes on a heartfelt note, describing how Michael’s wife designed the O Llama mascot—crafted to be simple, lovable, and easily recognizable, following principles of great logo design <a class=\"medius-aisummary\" data-starttime=\"00:12:31\">00:12:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:32\">00:13:32</a>. The hosts conclude with appreciation for O Llama’s creative spirit and its dedication to making AI accessible, private, and delightful for developers everywhere.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"23:55 - 00:10","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:10:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/f6e2dbf7-2eac-48a8-ac0f-9c687945de59","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/f6e2dbf7-2eac-48a8-ac0f-9c687945de59","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/f6e2dbf7-2eac-48a8-ac0f-9c687945de59","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/f6e2dbf7-2eac-48a8-ac0f-9c687945de59","speakerIds":["1768245294609build26-1772238023541001lKAd","1768245294609build26-1778108772492001j4Zf","1768245294609build26-1776980112222001zvph"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/f6e2dbf7-2eac-48a8-ac0f-9c687945de59","ogTitle":"Coding and Personal Agents with Ollama","ogDescription":"A live walkthrough of Ollama launch, showing how developers connect local and cloud open models to real-world applications in a single step. See Copilot CLI and OpenClaw set up live, switch between local and cloud models and connect the agent to web search and vision.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772238023541001lKAd","RegistrantKey":"1768245294609build26-1772238023541001lKAd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776980112222001zvph","RegistrantKey":"1768245294609build26-1776980112222001zvph"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778108772492001j4Zf","RegistrantKey":"1768245294609build26-1778108772492001j4Zf"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"cb1ead8e-0784-4d2b-a6ad-a6302d89f8d4","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pierce Boggan, Burke Holland","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780002547660001QdBY","langLocale":"en-US","sessionId":"1780002547660001QdBY","sessionInstanceId":"1780002547660001QdBY","sessionCode":"LIVE149","sessionCodeNormalized":"EN-US-LIVE149","title":"Day 1 Recap: What You Missed (or Watched Twice)","sortTitle":"day 1 recap: what you missed (or watched twice)","description":"Yesterday packed a lot in. We hit the biggest announcements, best demos, and moments worth coming back to before day two gets going.","aiDescription":"<p><strong>Opening and Introductions:</strong> The session begins with a warm welcome to day two of Microsoft Build by Burke, who lightheartedly encourages audience participation and energy <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>. He introduces himself and his colleague Pierce, clarifying their roles within Microsoft. Pierce explains that he works on the Visual Studio Code (VS Code) team focusing on GitHub Copilot, Copilot CLI, and related apps <a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:29\">00:00:29</a>. Burke adds that he also contributes to Copilot and VS Code, creating YouTube videos and engaging with the developer community. The tone is informal and friendly, setting up a conversational discussion for the remainder of the session.</p>\n\n<p><strong>Announcements and Community Engagement:</strong> Following introductions, Burke outlines several announcements and event highlights for attendees <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>. He notes that Windows-related sessions will begin later in the day and acknowledges both in-person and online audiences watching through Microsoft Developer, GitHub, and VS Code’s social channels <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:14\">00:01:14</a>. They joke about the “live coding booth” on the Expo floor—a glass-enclosed space likened to a fish tank—where developers code live while attendees watch <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:35\">00:01:35</a>. Pierce amusingly reveals he will be “locked in” for a live coding session later, continuing the playful banter.</p>\n\n<p><strong>Sessions and Fun Events:</strong> Burke then promotes upcoming sessions including the “VS Code team’s AI adoption story” led by Pierce and Josh <a class=\"medius-aisummary\" data-starttime=\"00:01:47\">00:01:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a>. They explain that rather than being about integrating AI into the product, the talk focuses on how the VS Code team itself uses AI in daily work <a class=\"medius-aisummary\" data-starttime=\"00:02:04\">00:02:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>. Next, they preview the “Copilot Game Show” happening on the keynote stage, co-hosted by Kent Dodds, where participants will attempt fast-paced live coding with Copilot in just 45 minutes <a class=\"medius-aisummary\" data-starttime=\"00:02:18\">00:02:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:42\">00:02:42</a>. They emphasize that the event will be entertaining and interactive, inviting attendees to join. Additionally, Burke notes the monthly VS Code release live streams as a great way for users to keep up with continuous updates, joking that even those working on the product struggle to track all the changes <a class=\"medius-aisummary\" data-starttime=\"00:02:43\">00:02:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:09\">00:03:09</a>.</p>\n\n<p><strong>Feature Highlights from Day One:</strong> Moving into technical discussions, Pierce and Burke recap key highlights from Build day one <a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>. They begin with the new GitHub Copilot App, describing it as a native desktop experience designed to manage development “agents” and workflows rather than being an editor or CLI tool <a class=\"medius-aisummary\" data-starttime=\"00:03:27\">00:03:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>. Pierce explains that unlike traditional tools managing single sessions, this app handles multiple concurrent agents, helping developers triage tasks, perform builds, and complete workflows efficiently <a class=\"medius-aisummary\" data-starttime=\"00:03:52\">00:03:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:10\">00:04:10</a>. They discuss real-world usability, suggesting that while some power users run numerous agents simultaneously, most developers can reasonably handle only three or four before losing track of tasks <a class=\"medius-aisummary\" data-starttime=\"00:04:35\">00:04:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:14\">00:05:14</a>. The focus, Pierce adds, should be outcome-driven rather than performative productivity—whether the tool achieves the intended results <a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:46\">00:05:46</a>.</p>\n\n<p><strong>Copilot SDK and Developer Tools:</strong> The conversation then turns to the GitHub Copilot SDK <a class=\"medius-aisummary\" data-starttime=\"00:06:11\">00:06:11</a>. Pierce clarifies that the SDK exposes the “agent loop” — the logic powering prompts, tools, and context within all Copilot products — allowing developers to integrate those capabilities into their own applications <a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:47\">00:06:47</a>. This lets teams package the Copilot runtime and experiment with building customized assistants. They encourage attendees with Copilot subscriptions to try one-shot experiments using the SDK <a class=\"medius-aisummary\" data-starttime=\"00:07:03\">00:07:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:16\">00:07:16</a>. The presenters note that few developers yet use open-source agents like Hermes or OpenClaw, but this SDK could expand experimentation in that space. The section underscores Microsoft’s focus on enabling developers to personalize and extend Copilot integration across tools.</p>\n\n<p><strong>New Coding Model and Closing Remarks:</strong> The final topic introduces Microsoft’s new coding model named “Mai Code 1 Fast” <a class=\"medius-aisummary\" data-starttime=\"00:07:20\">00:07:20</a>. Pierce explains that it is the first coding model built entirely in-house at Microsoft, optimized for GitHub Copilot <a class=\"medius-aisummary\" data-starttime=\"00:07:32\">00:07:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:49\">00:07:49</a>. Trained on Copilot’s internal tools and agent trajectories, it offers high efficiency and adaptive responses based on prompts <a class=\"medius-aisummary\" data-starttime=\"00:07:58\">00:07:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:11\">00:08:11</a>. The team expresses excitement about its rollout to VS Code developers, encouraging feedback as adoption expands through the model picker and Auto features <a class=\"medius-aisummary\" data-starttime=\"00:08:42\">00:08:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:51\">00:08:51</a>. Burke closes the session by thanking Pierce, inviting applause, and reminding attendees to catch his breakout session <a class=\"medius-aisummary\" data-starttime=\"00:08:52\">00:08:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>, ending on an upbeat, collegial note.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4b5c196a-6068-418a-81e9-a122f5107c15","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4b5c196a-6068-418a-81e9-a122f5107c15","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4b5c196a-6068-418a-81e9-a122f5107c15","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4b5c196a-6068-418a-81e9-a122f5107c15","speakerIds":["1768245294609build26-1775704791363001hLZq","1768245294609build26-1776347378057001Rvny"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4b5c196a-6068-418a-81e9-a122f5107c15","ogTitle":"Day 1 Recap: What You Missed (or Watched Twice)","ogDescription":"Yesterday packed a lot in. We hit the biggest announcements, best demos, and moments worth coming back to before day two gets going.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775704791363001hLZq","RegistrantKey":"1768245294609build26-1775704791363001hLZq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/LiveStage","backpackLink":null,"label":"Five hours of VS Code and GitHub, live at Microsoft Build. June 3, 9 AM to 2:10 PM","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"ac427fc4-c8a5-4fb8-a201-fe455ecdaebb","startDateTime":"2026-06-03T20:25:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Joanna Oikawa","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780009251275001cLSx","langLocale":"en-US","sessionId":"1780009251275001cLSx","sessionInstanceId":"1780009251275001cLSx","sessionCode":"LIVE156","sessionCodeNormalized":"EN-US-LIVE156","title":"Designing VS Code’s UX for the Agentic Era","sortTitle":"designing vs code’s ux for the agentic era","description":"Joanna will pull back the curtain on how VS Code's design team is quietly retooling the editor for the agentic era — not with a big rewrite, but with hundreds of small, deliberate UX shifts. She'll share real examples, the tradeoffs behind them, and a few things her team got wrong along the way. It's a candid look at evolving a product millions of developers already love.","aiDescription":"<p><strong>Introduction and Team Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>, the host welcomes viewers and introduces Joanna, the design lead for Visual Studio Code (VS Code). Joanna explains that VS Code’s design and UX engineering group is a small but powerful team of around forty-four designers. The discussion highlights how design and UX work hand-in-hand with engineering, enabling rapid iteration and unified product ownership. Joanna describes the vibrant UX design channel within the team, where prototypes and experiments often never reach production but inspire creative discussions among designers and engineers. This collaborative atmosphere allows ideas to emerge quickly and efficiently, with UX deeply integrated across the engineering culture.</p>\n\n<p><strong>Design Evolution and Decision Challenges:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:18\">00:02:18</a>, Joanna walks through examples of prototype work around the development of the “Agents Window,” describing how the team iterated on multiple designs in Figma starting back in December. They experimented with search menus, full-screen interfaces, and different approaches to display information across multiple workspaces. The conversation at <a class=\"medius-aisummary\" data-starttime=\"00:02:45\">00:02:45</a> reveals that these decisions are informed by user research and telemetry, focusing on how developers engage with sessions and projects. Joanna underscores the complexity of designing for evolving user modalities and the challenge of balancing existing patterns with new workflows. Rather than having one person make a final call, the team places early prototypes in Insider builds and continuously updates based on feedback, refining the product over time until it feels right.</p>\n\n<p><strong>Design and Development Integration:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:04:30\">00:04:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:11\">00:06:11</a>, Joanna discusses how VS Code’s design practices have advanced through tighter integration between Figma and actual code. With modern tools like Figma plugins and Microsoft’s component system, design elements move seamlessly between the design environment and the codebase. Designers actively submit pull requests, contributing directly to the product’s source code—a transformation from earlier workflows. Joanna shows prototypes of experimental themes and color refreshes, explaining that even if some concepts don’t ship, interactive prototypes make it easier to solicit feedback from engineers. Details such as panel shadows and the flashing update icon demonstrate how small visual details evolve through iterative refinement based on user responses.</p>\n\n<p><strong>Shifting Roles and AI-Enabled Breadth:</strong> The discussion around <a class=\"medius-aisummary\" data-starttime=\"00:06:27\">00:06:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:19\">00:08:19</a> explores the changing nature of team roles. Joanna agrees that the boundaries between designers, engineers, and product managers have blurred—everyone now contributes code, and everyone considers user experience. She introduces the concept of “depth versus breadth”: while her depth is in UX and visual design, AI tools now enable her to expand her breadth into areas like analytics and telemetry. Using Microsoft’s Kusto query language, she can efficiently analyze user data and make informed design decisions without depending entirely on specialized data analysts. This represents a shift toward multidimensional roles powered by AI, allowing team members to operate across traditional discipline boundaries while maintaining their core expertise.</p>\n\n<p><strong>Design Taste, Validation, and Continuous Improvement:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:08:56\">00:08:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>, the conversation turns philosophical. The host jokes about “AI psychosis” and the illusion of taste created by AI-generated visuals, emphasizing that authentic design requires human judgment. Joanna agrees, noting that deep experience and feedback loops remain critical for refining taste and ensuring products solve real user problems. She stresses the importance of validation—testing whether quick fixes and new features actually serve their purpose rather than just adding noise. By holding themselves accountable for removing features that fail or distract users, the team maintains VS Code’s reputation for clean and effective design even amid complex technology shifts and rapid AI-driven development.</p>\n\n<p><strong>Focus, Balance, and Conclusion:</strong> In the final segment from <a class=\"medius-aisummary\" data-starttime=\"00:11:28\">00:11:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:55\">00:12:55</a>, Joanna explains how the “Agents Window” design embodies restraint and intentional focus. The team reevaluated every element on screen to ensure it enhanced clarity rather than distraction, prioritizing navigation and multi-agent management while preventing cognitive overload. This iterative refinement highlights the principle that good design sometimes means pulling back, not adding more. The host closes by expressing appreciation for the VS Code design team and thanking Joanna for her insights. The segment wraps with applause and a lighthearted transition to the next video topic, leaving viewers with a deeper appreciation of the collaboration, craft, and ongoing evolution behind VS Code’s user experience.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"20:25 - 20:40","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:40:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/705403fe-f8ee-43bd-a13d-b49502401ad9","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/705403fe-f8ee-43bd-a13d-b49502401ad9","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/705403fe-f8ee-43bd-a13d-b49502401ad9","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/705403fe-f8ee-43bd-a13d-b49502401ad9","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1778714376145001v89o"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/705403fe-f8ee-43bd-a13d-b49502401ad9","ogTitle":"Designing VS Code’s UX for the Agentic Era","ogDescription":"Joanna will pull back the curtain on how VS Code's design team is quietly retooling the editor for the agentic era — not with a big rewrite, but with hundreds of small, deliberate UX shifts. She'll share real examples, the tradeoffs behind them, and a few things her team got wrong along the way. It's a candid look at evolving a product millions of developers already love.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778714376145001v89o","RegistrantKey":"1768245294609build26-1778714376145001v89o"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/Release","backpackLink":null,"label":"Stay up to date with VS Code releases","category":"Web link"},{"link":"https://aka.ms/VSCode/DesignYT","backpackLink":null,"label":"How VS Code designers are getting closer to the codebase using agents and VS Code","category":"Web link"},{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Build your VS Code and Copilot skills with free, hands-on courses","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"dbb2f296-2529-42e3-89f0-bdf573fbb6e3","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Patrick Nikoletich","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780003482697001njSw","langLocale":"en-US","sessionId":"1780003482697001njSw","sessionInstanceId":"1780003482697001njSw","sessionCode":"LIVE151","sessionCodeNormalized":"EN-US-LIVE151","title":"Extend GitHub Copilot with the Copilot SDK","sortTitle":"extend github copilot with the copilot sdk","description":"The Copilot SDK lets you extend, customize, and build your own agentic experiences using the same runtime that powers GitHub Copilot. See it in action.","aiDescription":"<p><strong>Introduction and Event Context:</strong> The video opens with the host greeting viewers from Build 2026 and introducing Patrick Nicoletich, a lead on the Copilot SDK team <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>. They share a lighthearted conversation about the event’s large venue before transitioning to the topic of the Copilot SDK. Patrick notes the SDK’s general availability (GA) status, announced just the day before, marking a major milestone for developers across Microsoft and its subsidiaries like LinkedIn, Xbox, and GitHub. The SDK’s 1.0 release represents a stable, production-ready version that supports developers in integrating Copilot capabilities into various applications <a class=\"medius-aisummary\" data-starttime=\"00:00:59\">00:00:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>.</p>\n\n<p><strong>SDK Launch and Technical Overview:</strong> Patrick explains that the Copilot SDK, first introduced in technical preview, now supports six languages including Rust and Java <a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:36\">00:02:36</a>. The SDK functions as a lightweight client over the Copilot CLI server, enabling developers to embed the Copilot experience directly into custom software. It extends the same infrastructure powering GitHub’s Copilot into broader applications. Ongoing improvements focus on performance, including a rewrite of the runtime in Rust to reduce memory usage from approximately 120MB to under 10MB while increasing portability across environments <a class=\"medius-aisummary\" data-starttime=\"00:03:11\">00:03:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:42\">00:03:42</a>.</p>\n\n<p><strong>Performance Evolution and Engineering Goals:</strong> Discussion turns to internal testing that highlights clear performance improvements from the Rust runtime, showing better efficiency and scalability across multiple devices <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:05\">00:04:05</a>. The speakers note how GitHub teams embraced experimental approaches to evaluate these changes, using large-scale model interactions to push boundaries and develop adaptable, “agentic” workflows. They reflect on how advancements in Copilot have redefined development culture—encouraging exploration, reducing friction between idea and outcome, and freeing teams from computational constraints like limited tokens <a class=\"medius-aisummary\" data-starttime=\"00:04:44\">00:04:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:20\">00:05:20</a>.</p>\n\n<p><strong>Demonstration of “Whim” Personal Assistant:</strong> For a practical example, Patrick demonstrates his own Copilot SDK-based app, called “Whim,” designed as a personal productivity assistant for PM work <a class=\"medius-aisummary\" data-starttime=\"00:05:46\">00:05:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:16\">00:06:16</a>. The app uses a canvas-based interface instead of traditional chat interactions, allowing users to spawn, manage, and track agents visually. He shows how Whim automates writing, issue triage, and project management tasks through agents deployed in ephemeral cloud environments or secure local sandboxes. These agents operate autonomously but transparently, updating a shared workspace as progress unfolds <a class=\"medius-aisummary\" data-starttime=\"00:08:07\">00:08:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:25\">00:09:25</a>. The demonstration highlights how multi-agent orchestration can blend seamlessly into everyday workflows.</p>\n\n<p><strong>Adoption, Usability, and Broader Reflections:</strong> The conversation moves toward how real-world developers can use the SDK to build personal tools instead of relying on generalized AI systems <a class=\"medius-aisummary\" data-starttime=\"00:11:11\">00:11:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:45\">00:11:45</a>. Patrick emphasizes that grounding experiments in clear, communicable use cases demystifies AI and makes it more practical. New SDK features like remote access, ephemeral cloud sandboxes, and local sandbox environments are now available for preview, offering developers flexibility to prototype without heavy setup <a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:19\">00:12:19</a>. Both speakers encourage viewers to start small—automating daily tasks, learning model behavior, and evolving trust with their AI agents over time.</p>\n\n<p><strong>Conclusion and Future Sessions:</strong> In concluding remarks, they discuss their preferred models—GPT 5.5 for implementation tasks and Opus for creativity—describing their complementary strengths <a class=\"medius-aisummary\" data-starttime=\"00:14:12\">00:14:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:03\">00:15:03</a>. Patrick invites attendees to his upcoming session from 4:00–5:00 PM in Breakout 1 for a deeper technical dive into the Copilot SDK’s future and cross-Microsoft integration <a class=\"medius-aisummary\" data-starttime=\"00:15:08\">00:15:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:25\">00:15:25</a>. The host wraps up by reaffirming the transformative potential of building tailored AI workflows, thanking Patrick for the insights and promising viewers more live conversations ahead.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/175f1178-984b-4d72-b15b-108d8d67dec2","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/175f1178-984b-4d72-b15b-108d8d67dec2","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/175f1178-984b-4d72-b15b-108d8d67dec2","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/175f1178-984b-4d72-b15b-108d8d67dec2","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1777003136185001uvnL"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/175f1178-984b-4d72-b15b-108d8d67dec2","ogTitle":"Extend GitHub Copilot with the Copilot SDK","ogDescription":"The Copilot SDK lets you extend, customize, and build your own agentic experiences using the same runtime that powers GitHub Copilot. See it in action.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777003136185001uvnL","RegistrantKey":"1768245294609build26-1777003136185001uvnL"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/GHCP/SDK","backpackLink":null,"label":"GitHub Copilot SDK","category":"Web link"},{"link":"https://aka.ms/GHCP/SDK-docs","backpackLink":null,"label":"Get started","category":"Web link"},{"link":"https://aka.ms/AwesomeCopilot","backpackLink":null,"label":"Examples and recipes","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"4a595f04-505a-46ba-a141-7a6aa4f13545","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vivek Chauhan","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213039221001LFK7","langLocale":"en-US","sessionId":"1778213039221001LFK7","sessionInstanceId":"1778213039221001LFK7","sessionCode":"LIVESP125","sessionCodeNormalized":"EN-US-LIVESP125","title":"Fireworks AI quick take: Live from Microsoft Build","sortTitle":"fireworks ai quick take: live from microsoft build","description":"Get a sneak peek of the latest from Fireworks AI featuring Microsoft’s Karuana Gatimu and Vivek Chauhan","aiDescription":"<p><strong>Introduction from the Microsoft Build Show Floor:</strong> The video opens with the host welcoming viewers back to the show floor at Microsoft Build in San Francisco (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a>). The host introduces the Fireworks AI booth and expresses excitement about showcasing new innovations. Vivek, a representative from Fireworks AI, joins the conversation to discuss the company’s platform and its impact on artificial intelligence production environments (<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>).</p>\n\n<p><strong>Overview of Fireworks AI’s Technology:</strong> Vivek explains that Fireworks AI is a “production AI platform for open weight models,” designed to handle extensive real-world AI workloads (<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>). The platform processes over 30 trillion tokens each day and supports more than 10,000 enterprise customers globally, including major organizations like Cursor and Uber (<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>). These clients use Fireworks AI not only for inference but also to fine-tune models, embedding their proprietary data directly within applications. Vivek also references RL (reinforcement learning) capabilities used by products such as Cursor’s Composer 2 and 2.5 to scale their AI systems efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>).</p>\n\n<p><strong>Microsoft and Fireworks AI Partnership:</strong> The discussion transitions to the strategic partnership between Microsoft and Fireworks AI (<a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>). Vivek details how customers can now access Fireworks AI’s large-scale inference services directly through Azure Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:13\">00:01:13</a>). This integration allows users to benefit from Azure’s governance, reliability, and security while also enjoying simplified billing. Moreover, developers gain access to a broader selection of open weight models and can train on Fireworks AI while running inference seamlessly on Azure, bringing together the strengths of both ecosystems (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:31\">00:01:31</a>).</p>\n\n<p><strong>Event Participation and Excitement:</strong> The host notes that this is Vivek’s first time attending Microsoft Build (<a class=\"medius-aisummary\" data-starttime=\"00:01:40\">00:01:40</a>). Vivek shares his enthusiasm about meeting fellow innovators and presenting at the event alongside his colleagues, who have scheduled talks later in the week (<a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>). The positive tone reflects both excitement about the technology on display and the collaborative atmosphere of the conference.</p>\n\n<p><strong>Conclusion and Viewer Engagement:</strong> In closing, the host invites viewers to explore Fireworks AI demos available on their sponsor showcase page and encourages the digital audience to access links to sessions presented by Vivek and his team (<a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>). The video ends with a note of appreciation to Vivek for joining and an invitation for viewers to stay tuned for more insights from Microsoft Build (<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b983d7e1-2c7f-4ee0-9c08-537ea501e895","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b983d7e1-2c7f-4ee0-9c08-537ea501e895","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b983d7e1-2c7f-4ee0-9c08-537ea501e895","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b983d7e1-2c7f-4ee0-9c08-537ea501e895","speakerIds":["1768245294609build26-1776288992886001q5Ev"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.772Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Fireworks AI"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b983d7e1-2c7f-4ee0-9c08-537ea501e895","ogTitle":"Fireworks AI quick take: Live from Microsoft Build","ogDescription":"Get a sneak peek of the latest from Fireworks AI featuring Microsoft’s Karuana Gatimu and Vivek Chauhan","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776288992886001q5Ev","RegistrantKey":"1768245294609build26-1776288992886001q5Ev"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"fc0a2923-a667-4439-9824-1cc2c8a87cab","startDateTime":"2026-06-03T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Seth Juarez","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780007941340001jVRl","langLocale":"en-US","sessionId":"1780007941340001jVRl","sessionInstanceId":"1780007941340001jVRl","sessionCode":"LIVE162","sessionCodeNormalized":"EN-US-LIVE162","title":"From issue to merge in one loop: the GitHub Copilot app","sortTitle":"from issue to merge in one loop: the github copilot app","description":"What if you could hand off an issue, watch agents work it in real time, review the diff, and merge, all without leaving one screen? The GitHub Copilot app is a new desktop experience built for agent-driven development.","aiDescription":"<p><strong>Opening Banter and Setup:</strong> The video begins with lighthearted chatter among the hosts as they prepare to go live. There is a casual exchange about stale doughnuts and the lack of lunch invitations, setting an informal tone (<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a>). The hosts joke about not being invited to community events or GitHub meetups, and one recounts showing up to an MVP gathering without being officially invited (<a class=\"medius-aisummary\" data-starttime=\"00:00:49\">00:00:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>). This playful beginning transitions into formal discussion topics where Seth introduces that they have “official business” to cover and will talk about GitHub Copilot and agents (<a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>).</p>\n\n<p><strong>Introducing GitHub Copilot and Workflow Contexts:</strong> The hosts shift to explaining how GitHub Copilot fits into different developer workflows. They categorize the development tools into the command line interface (CLI), Visual Studio Code, and the GitHub Copilot app (<a class=\"medius-aisummary\" data-starttime=\"00:02:10\">00:02:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>). Seth explains that each tool aligns with a specific level of focus — CLI for working on one isolated task, VS Code for project-level attention and seeing code details, and the Copilot app for broad, multi-project context (<a class=\"medius-aisummary\" data-starttime=\"00:02:32\">00:02:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:50\">00:03:50</a>). He demonstrates how the Copilot app acts as a “click stop up,” showing multiple repositories at once and giving developers a way to start their mornings with an overview of pending tasks and security issues (<a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>).</p>\n\n<p><strong>Holistic Decision-Making and Ideation with Copilot:</strong> The conversation deepens as Seth describes code as “the easy part” — emphasizing that the real complexity is deciding what needs to be done and when. He demonstrates how the GitHub Copilot app supports that level of decision-making by giving a holistic overview of projects, enabling users to see releases, dependencies, and tasks before focusing in on writing code (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:04\">00:06:04</a>). The app filters to only what the user is working on, unlike GitHub’s broad notification system, helping streamline task management (<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:09\">00:07:09</a>). They highlight “quick chat,” a tool in the app that supports ideation outside a specific project, where ideas can evolve without changing files, before being intentionally connected to a concrete project later (<a class=\"medius-aisummary\" data-starttime=\"00:07:12\">00:07:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>).</p>\n\n<p><strong>App Integration and Development Structure:</strong> Seth explains that he builds various helper apps — such as CutReady, Snipse, and Allusum — which support presentations and workflow automation using technologies like Tauri and Rust. He humorously mispronounces “Tauri,” clarifies its nature as a lightweight web app framework, and discusses using shared infrastructure among these tools (<a class=\"medius-aisummary\" data-starttime=\"00:09:19\">00:09:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:52\">00:09:52</a>). By leveraging GitHub Copilot’s “quick chat,” he identifies repeating patterns and explores consolidating code for efficiency across projects (<a class=\"medius-aisummary\" data-starttime=\"00:10:02\">00:10:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:47\">00:10:47</a>). The team discusses how they can connect ideation sessions to real code bases once decisions become actionable, comparing this dynamic to how their dinner plans would be “disconnected until connected” (<a class=\"medius-aisummary\" data-starttime=\"00:10:47\">00:10:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:59\">00:10:59</a>).</p>\n\n<p><strong>Key Productivity Enhancements and Canvas Feature:</strong> Transitioning to new features, Seth praises the Copilot app’s project onboarding simplicity — being able to instantly open GitHub repositories without manually cloning or configuring work trees (<a class=\"medius-aisummary\" data-starttime=\"00:11:20\">00:11:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:46\">00:11:46</a>). He calls this “the single best feature of any app.” With minutes remaining, the hosts discuss the “Canvas” feature, which introduces a new way to interact visually with large language models beyond text chat (<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:20\">00:12:20</a>). Canvas lets users build interactive diagrams with their agents, explore repositories, inspect components, and visualize divergences among branches or project dependencies. Seth describes using this in his Elusum app to click visual elements and prompt explanations directly from the agent (<a class=\"medius-aisummary\" data-starttime=\"00:12:29\">00:12:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:48\">00:13:48</a>).</p>\n\n<p><strong>Closing Remarks and Humor:</strong> The session wraps with reflections on how rapidly and meaningfully AI and developer tools are evolving. They celebrate simplicity and accessibility, joking that everything now moves “so slowly and it’s so easy to understand all the new features” (<a class=\"medius-aisummary\" data-starttime=\"00:14:01\">00:14:01</a>). The hosts end with laughter, light jokes about exhaustion, and farewells, thanking each other and the audience before signing off and promising to be “right back” (<a class=\"medius-aisummary\" data-starttime=\"00:14:07\">00:14:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:15\">00:14:15</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"19:30 - 19:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4ab34223-dd67-4829-846d-ca53fe37892b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4ab34223-dd67-4829-846d-ca53fe37892b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4ab34223-dd67-4829-846d-ca53fe37892b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4ab34223-dd67-4829-846d-ca53fe37892b","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1751379013075001HHpX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4ab34223-dd67-4829-846d-ca53fe37892b","ogTitle":"From issue to merge in one loop: the GitHub Copilot app","ogDescription":"What if you could hand off an issue, watch agents work it in real time, review the diff, and merge, all without leaving one screen? The GitHub Copilot app is a new desktop experience built for agent-driven development.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/GitHubCopilot/app","backpackLink":null,"label":"Get the GitHub Copilot app","category":""},{"link":"https://aka.ms/GitHubCopilot/app-docs","backpackLink":null,"label":"Read the GitHub Copilot app docs","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d8777250-dbeb-4a2e-acda-d926ad5090f4","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Adrian Macias","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213499691001IPsV","langLocale":"en-US","sessionId":"1778213499691001IPsV","sessionInstanceId":"1778213499691001IPsV","sessionCode":"LIVESP128","sessionCodeNormalized":"EN-US-LIVESP128","title":"From local AI PCs to Azure: The future of open-source AI development","sortTitle":"from local ai pcs to azure: the future of open-source ai development","description":"AI development workflows are changing rapidly as developers experiment with agentic AI, AI-assisted coding, and increasingly flexible deployment strategies. In this Microsoft Build conversation, AMD and Microsoft discuss how open-source AI ecosystems, Ryzen AI PCs, ROCm, and Azure infrastructure are enabling developers to experiment, adapt, and scale AI workloads with greater flexibility across evolving AI environments.","aiDescription":"<p><strong>Introduction and Setting:</strong> The video opens with a welcome back to the broadcast stage at Microsoft Build in the vibrant city of San Francisco <a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>. The host sets the tone by highlighting the event’s focus on innovation at scale and speed and introduces Adrian, a returning guest known for his insights on developer experiences. The conversation begins with an emphasis on helping developers better understand what’s changing for them and providing actionable guidance <a class=\"medius-aisummary\" data-starttime=\"00:00:32\">00:00:32</a>.</p>\n\n<p><strong>Transformation of the Developer Experience:</strong> Adrian discusses the rapid acceleration of innovation over the past six months, noting that developers are experimenting in areas they previously couldn’t access <a class=\"medius-aisummary\" data-starttime=\"00:00:51\">00:00:51</a>. He identifies a convergence of technologies—software, hardware, and agentic AI—that is altering how problems are solved. Rather than focusing on deep expertise, developers now thrive as experimenters who embrace risk and novel solutions <a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>. The reduced cost to explore new ideas, Adrian explains, powers this sense of freedom and opens vast possibilities for innovation <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>.</p>\n\n<p><strong>AI and Machine Transformation:</strong> A key idea in the discussion is that “the machine has fundamentally changed,” referring to how AI and especially agentic AI technologies transform coding and design flows <a class=\"medius-aisummary\" data-starttime=\"00:02:43\">00:02:43</a>. Adrian describes this shift as empowering developers to take more risks and iterate faster without penalty. The conversation draws parallels between creativity in engineering and “happy accidents” in art, as unexpected outcomes often lead to breakthroughs. He stresses the importance of fostering open ecosystems—open software and platforms—to support these innovations, which both AMD and Microsoft are investing heavily in <a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>.</p>\n\n<p><strong>Exploring Modalities and Platforms:</strong> Adrian outlines how developers can experiment with different modalities and execution environments <a class=\"medius-aisummary\" data-starttime=\"00:04:39\">00:04:39</a>. Developers can now run workloads on cloud, device, or hybrid setups, enabling local or distributed AI computing. He elaborates on advancements such as AI PCs with neural processing units <a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a> and expands on how multi-modal AI—combining speech, text, vision, and emotional recognition—creates more interactive experiences. The conversation encourages developers to join showcases and use online Microsoft Build sessions to deepen learning, emphasizing that these resources are freely available to a global audience <a class=\"medius-aisummary\" data-starttime=\"00:06:17\">00:06:17</a>.</p>\n\n<p><strong>Future of Agentic AI and Developer Choices:</strong> Looking ahead three to six months, Adrian predicts that agentic AI microservices will drive applications to become dynamic, adaptive, and personalized <a class=\"medius-aisummary\" data-starttime=\"00:07:06\">00:07:06</a>. Developers will increasingly consider orchestration among multiple intelligent agents interacting across systems <a class=\"medius-aisummary\" data-starttime=\"00:07:37\">00:07:37</a>. He explains how these systems make intelligent decisions based on cost, resource constraints, and quality, even determining whether to run tasks locally or in the cloud <a class=\"medius-aisummary\" data-starttime=\"00:08:24\">00:08:24</a>. Security plays a crucial role in this new landscape, with modern routing techniques ensuring personal data remains protected while still leveraging cloud AI foundations <a class=\"medius-aisummary\" data-starttime=\"00:09:25\">00:09:25</a>.</p>\n\n<p><strong>Creativity, Announcements, and Conclusion:</strong> Adrian reflects on how these technological changes unleash creativity in developers, stating that innovation is directly linked to iteration speed <a class=\"medius-aisummary\" data-starttime=\"00:10:20\">00:10:20</a>. The discussion highlights courage in experimentation and the partnership between Microsoft and AMD on the Win MLSDK project, which democratizes machine learning access and reduces development friction <a class=\"medius-aisummary\" data-starttime=\"00:11:07\">00:11:07</a>. They also explore evolving ideas of quality in AI—from simple accuracy to complex behaviors of intelligent agents <a class=\"medius-aisummary\" data-starttime=\"00:12:17\">00:12:17</a>. The session wraps up with gratitude, encouragement to visit AMD’s showcase and the Microsoft Build site for demos and resources, and a closing thank you to the audience <a class=\"medius-aisummary\" data-starttime=\"00:13:25\">00:13:25</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:30 - 21:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4f44c7a4-7e1a-4fc3-b614-690604749d97","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4f44c7a4-7e1a-4fc3-b614-690604749d97","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4f44c7a4-7e1a-4fc3-b614-690604749d97","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4f44c7a4-7e1a-4fc3-b614-690604749d97","speakerIds":["1768245294609build26-1776355971597001QUS8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.773Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4f44c7a4-7e1a-4fc3-b614-690604749d97","ogTitle":"From local AI PCs to Azure: The future of open-source AI development","ogDescription":"AI development workflows are changing rapidly as developers experiment with agentic AI, AI-assisted coding, and increasingly flexible deployment strategies. In this Microsoft Build conversation, AMD and Microsoft discuss how open-source AI ecosystems, Ryzen AI PCs, ROCm, and Azure infrastructure are enabling developers to experiment, adapt, and scale AI workloads with greater flexibility across evolving AI environments.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776355971597001QUS8","RegistrantKey":"1768245294609build26-1776355971597001QUS8"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d8437000-40f7-4c54-b3b2-5815ad6b444d","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779306381282001lPnM","langLocale":"en-US","sessionId":"1779306381282001lPnM","sessionInstanceId":"1779306381282001lPnM","sessionCode":"LIVE104","sessionCodeNormalized":"EN-US-LIVE104","title":"GitHub, Copilot, VS Code, and More: Live from San Francisco","sortTitle":"github, copilot, vs code, and more: live from san francisco","description":"Join us on stage for live demos, surprise guests, games, live coding, and more. The latest from GitHub, Copilot, VS Code, Visual Studio, and friends, straight from the teams building it.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"16:00 - 17:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:55:00Z","durationInMinutes":115,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/01fc607c-146a-4fd1-b261-04bf01a9b281","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/01fc607c-146a-4fd1-b261-04bf01a9b281","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/01fc607c-146a-4fd1-b261-04bf01a9b281","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/01fc607c-146a-4fd1-b261-04bf01a9b281","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.235Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/01fc607c-146a-4fd1-b261-04bf01a9b281","ogTitle":"GitHub, Copilot, VS Code, and More: Live from San Francisco","ogDescription":"Join us on stage for live demos, surprise guests, games, live coding, and more. The latest from GitHub, Copilot, VS Code, Visual Studio, and friends, straight from the teams building it.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"f28dee03-33a3-4e34-81eb-49f329e75d33","startDateTime":"2026-06-03T21:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeff Hollan, Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780010316953001vqy2","langLocale":"en-US","sessionId":"1780010316953001vqy2","sessionInstanceId":"1780010316953001vqy2","sessionCode":"LIVE170","sessionCodeNormalized":"EN-US-LIVE170","title":"Hosted Agents Live: From Code to Production AI at Scale","sortTitle":"hosted agents live: from code to production ai at scale","description":"Join us live from Build as we break down how developers are moving from coding agents to running production‑grade AI systems using hosted agents in Microsoft Foundry. We’ll show how to go from local prototyping to secure, scalable deployment—covering identity, isolation, evaluation, and lifecycle management—so you can ship real‑world agents with confidence.","aiDescription":"<p><strong>Introduction and Roles:</strong> The video begins with greetings from Lee Star, Principal Cloud Advocate at Microsoft, who introduces his guest Jeff Holland, Partner Director for Foundry Agent <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>. Jeff describes his responsibilities in Foundry's product management team, particularly focusing on enabling developers to build and deploy cloud-scale AI agents <a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>. He notes that his team supports builders with selecting models, hosting agents, and managing all necessary tools and evaluation components.</p>\n\n<p><strong>Hosted Agents Overview:</strong> The discussion moves to Hosted Agents, a recently launched preview product within Foundry <a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>. Jeff explains that unlike traditional prompt-based agents, Hosted Agents allow developers to bring their own agent code or harness. Users can leverage frameworks like Lang Graph, Microsoft Agent Framework, Anthropic Agent SDK, and GitHub Copilot SDK to define and test agents locally before Foundry hosts them in the cloud <a class=\"medius-aisummary\" data-starttime=\"00:01:09\">00:01:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>. This system connects agent logic to models, observability, and tooling to enable scalable deployment without sacrificing developer control.</p>\n\n<p><strong>From Prototype to Production:</strong> Jeff highlights the challenge of moving from prototype agents—quickly created with tools like Copilot—to production-ready reliable agents <a class=\"medius-aisummary\" data-starttime=\"00:02:11\">00:02:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:28\">00:03:28</a>. Quality assurance is critical, with enterprises demanding accurate responses and strong evaluation metrics. He mentions observability, lifecycle management, versioning, and scalability as core considerations. The transition mirrors classical DevOps patterns with similar requirements for monitoring, testing, and continuous improvement <a class=\"medius-aisummary\" data-starttime=\"00:03:47\">00:03:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>.</p>\n\n<p><strong>Identity, Security, and Governance:</strong> Addressing one of the biggest developer challenges—agent identity and isolation—Jeff describes how Microsoft integrates robust identity management through Intra’s Agent Identity <a class=\"medius-aisummary\" data-starttime=\"00:04:12\">00:04:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>. These identities enable both standard managed authentication and delegated “on-behalf-of” access patterns, allowing secure interaction with user data while maintaining telemetry tracking. Foundry automatically provisions agent identities, simplifying configuration for developers <a class=\"medius-aisummary\" data-starttime=\"00:05:16\">00:05:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:33\">00:05:33</a>. Jeff further elaborates on new Foundry role frameworks (user, contributor, observability) and efforts to define least-privileged access and guardrails for autonomous agents <a class=\"medius-aisummary\" data-starttime=\"00:05:58\">00:05:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:56\">00:06:56</a>.</p>\n\n<p><strong>Production Insights and Future Directions:</strong> Reflecting on developer adoption, Jeff warns that many underestimate the potential of well-designed agents <a class=\"medius-aisummary\" data-starttime=\"00:07:08\">00:07:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:49\">00:07:49</a>. He emphasizes balancing governance with autonomy, allowing agents sufficient flexibility to surprise users with useful outcomes. Modern agent harnesses like Open Claw, Hermes Agent, and GitHub Copilot CLI represent significant advances. By combining a strong harness with powerful models, developers can unlock unexpected value and innovation in production environments <a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:24\">00:08:24</a>.</p>\n\n<p><strong>Conclusion and Developer Call to Action:</strong> As the conversation concludes, Jeff encourages developers to explore Microsoft Foundry by following its documentation and quick-start guides <a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>. He recommends reading blogs and trying out hands-on labs to rapidly learn Hosted Agents’ capabilities. Lee thanks Jeff for the insights, closing the video with enthusiasm about the growing possibilities of hosted AI agents and the Foundry ecosystem. The overall message is that now is the perfect time for developers to experiment and shape their future agent-driven solutions <a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:20 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":10,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ad9a77fd-946a-42fd-aaf6-c2baccaf73ff","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ad9a77fd-946a-42fd-aaf6-c2baccaf73ff","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ad9a77fd-946a-42fd-aaf6-c2baccaf73ff","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ad9a77fd-946a-42fd-aaf6-c2baccaf73ff","speakerIds":["1768245294609build26-1774284996168001pgz1","1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ad9a77fd-946a-42fd-aaf6-c2baccaf73ff","ogTitle":"Hosted Agents Live: From Code to Production AI at Scale","ogDescription":"Join us live from Build as we break down how developers are moving from coding agents to running production‑grade AI systems using hosted agents in Microsoft Foundry. We’ll show how to go from local prototyping to secure, scalable deployment—covering identity, isolation, evaluation, and lifecycle management—so you can ship real‑world agents with confidence.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996168001pgz1","RegistrantKey":"1768245294609build26-1774284996168001pgz1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/foundry-hosted-agents","backpackLink":null,"label":"Hosted Agents Live: From Code to Production AI at Scale","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"95959599-8146-4407-8c73-6721560f2682","startDateTime":"2026-06-03T17:55:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pierce Boggan, Pengcheng He, Seth Juarez, Yang Liu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780005736167001eMpo","langLocale":"en-US","sessionId":"1780005736167001eMpo","sessionInstanceId":"1780005736167001eMpo","sessionCode":"LIVE158","sessionCodeNormalized":"EN-US-LIVE158","title":"How Microsoft AI builds coding models optimized for GitHub Copilot","sortTitle":"how microsoft ai builds coding models optimized for github copilot","description":"Go behind the scenes with Microsoft AI to learn how to build and optimize coding models for GitHub Copilot. This session will explore what makes code-focused models different—from training and evaluation to performance, safety, and real-world developer feedback. You’ll hear how Microsoft AI is advancing model quality for the workflows developers care about most, and how those innovations show up in GitHub Copilot experiences used by millions of developers.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video opens with a friendly welcome from the host at <a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>, expressing excitement about the AI innovations announced at Microsoft Build. He notes that seven models were released and highlights one that particularly interests him as a developer — \"Mai Code Flash.\" He introduces two researchers, Young and Hunter, who lead work on the model and explains that the conversation will explore what makes this code model unique, especially its integration with GitHub Copilot and developer workflows. This sets the stage for a technical deep dive into how it was built and optimized specifically for coding assistance in Visual Studio Code.</p>\n\n<p><strong>Purpose-Built Design for GitHub Copilot:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:22\">00:01:22</a>, the researchers discuss that Mai Code Flash was not adapted from a general-purpose model but designed from the ground up for GitHub Copilot usage. Peng explains that this approach enables it to understand developers’ intentions, handle context, and make targeted edits quickly. The team optimized it for real user workflows, ensuring smooth interactions in VS Code where the model can assist with programming tasks and understand coding context rather than just generating snippets. They emphasize that this intentional grounding in developer tooling makes it responsive and accurate within real coding environments.</p>\n\n<p><strong>Training Process and Reinforcement Learning:</strong> The video then moves into an explanation of how Mai Code Flash was trained at <a class=\"medius-aisummary\" data-starttime=\"00:02:46\">00:02:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>. The researchers describe a multi-stage process involving supervised fine-tuning followed by reinforcement learning. They start by teaching the model to follow user instructions on simple coding tasks and progressively increase complexity. Eventually, reinforcement learning aligns the model’s behavior through real interactions, using GitHub and VS Code environments as realistic training harnesses. This makes evaluation and production share the same workflow. The result is a model that behaves naturally for developers, improving as if it’s “going to school” — learning coding fundamentals before tackling advanced tasks. Customized benchmarks mirror real developer environments to verify performance during each stage.</p>\n\n<p><strong>Model Architecture and Efficiency:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:05:27\">00:05:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:20\">00:07:20</a>, the discussion shifts to the model’s architecture. The researchers explain that Mai Code Flash uses a Mixture of Experts (MoE) approach. While the total capacity reaches 137 billion parameters, only 5 billion are active per inference, providing high efficiency and reducing token usage. This structure achieves a balance between intelligence, speed, and cost. Each expert network handles relevant parts of a task, enabling fast computation and low latency. They highlight that token efficiency and smart activation make it \"a beast at its size\" — powerful yet lightweight for daily coding needs, allowing superior speed compared to similarly intelligent models.</p>\n\n<p><strong>Demonstrations and Real-World Application:</strong> In the middle section of the video at <a class=\"medius-aisummary\" data-starttime=\"00:09:20\">00:09:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:20\">00:11:20</a>, the team presents live demos of Mai Code Flash in VS Code. The first demo shows the model building a front-end web app within seconds, responding efficiently and generating clean results. A second demo features creation of a simple AI-based game, showcasing enhanced UI generation ability. They note that the model is being gradually rolled out to users and visible via the “model picker” in Copilot. Discussions follow on testing and continuous refinement through A/B experiments, ensuring user engagement and minimal cancellation during interaction. Results show the model performs comparably to large frontier models for most coding tasks, providing a fast, reliable developer experience.</p>\n\n<p><strong>Integration, Future, and Closing Thoughts:</strong> In the final portion, starting around <a class=\"medius-aisummary\" data-starttime=\"00:14:35\">00:14:35</a> and continuing to <a class=\"medius-aisummary\" data-starttime=\"00:16:15\">00:16:15</a>, they demonstrate how Mai Code Flash integrates with other Mai models — including Mai Transcribe and Mai Voice — enabling multi-agent collaborative coding environments where AI agents execute real development tasks. The researchers attribute this to natural synergy among the seven released Mai models. When asked what developers should take away, they emphasize imagination and continuous improvement. Users are encouraged to provide feedback to refine future “powerful beasts.” The host concludes enthusiastically, thanking the team and urging developers to stay tuned for upcoming iterations, applauding the innovation that merges efficiency, intelligence, and practical coding application.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"17:55 - 18:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:15:00Z","durationInMinutes":20,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0a23b038-c73a-4b01-901d-9098e033af9a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0a23b038-c73a-4b01-901d-9098e033af9a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0a23b038-c73a-4b01-901d-9098e033af9a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0a23b038-c73a-4b01-901d-9098e033af9a","speakerIds":["1768245294609build26-1775704791363001hLZq","1768245294609build26-1780270354181001qcgk","1768245294609build26-1751379013075001HHpX","1768245294609build26-1780079069034001L7yX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0a23b038-c73a-4b01-901d-9098e033af9a","ogTitle":"How Microsoft AI builds coding models optimized for GitHub Copilot","ogDescription":"Go behind the scenes with Microsoft AI to learn how to build and optimize coding models for GitHub Copilot. This session will explore what makes code-focused models different—from training and evaluation to performance, safety, and real-world developer feedback. You’ll hear how Microsoft AI is advancing model quality for the workflows developers care about most, and how those innovations show up in GitHub Copilot experiences used by millions of developers.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775704791363001hLZq","RegistrantKey":"1768245294609build26-1775704791363001hLZq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780079069034001L7yX","RegistrantKey":"1768245294609build26-1780079069034001L7yX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780270354181001qcgk","RegistrantKey":"1768245294609build26-1780270354181001qcgk"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/MAI/GHCP","backpackLink":null,"label":"Learn more about yesterday’s MAI launch","category":"Web link"},{"link":"https://aka.ms/GHCP/automodel","backpackLink":null,"label":"Learn more about GitHub Copilot auto model selection","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"83160112-eca4-4ae4-9dab-42608cfb5585","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Seth Juarez, Julia Kasper","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780005399135001xJby","langLocale":"en-US","sessionId":"1780005399135001xJby","sessionInstanceId":"1780005399135001xJby","sessionCode":"LIVE161","sessionCodeNormalized":"EN-US-LIVE161","title":"How we ship models in VS Code","sortTitle":"how we ship models in vs code","description":"Shipping the right AI model for each task requires a lot of testing and evaluation. Get an inside look at how the VS Code and Copilot teams assess model quality, decide when to roll out updates, and balance capability with reliability.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e408c0a9-ef09-4ef4-b991-9345a95fde70","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e408c0a9-ef09-4ef4-b991-9345a95fde70","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e408c0a9-ef09-4ef4-b991-9345a95fde70","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e408c0a9-ef09-4ef4-b991-9345a95fde70","speakerIds":["1768245294609build26-1751379013075001HHpX","1768245294609build26-1774303544298001acRa"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e408c0a9-ef09-4ef4-b991-9345a95fde70","ogTitle":"How we ship models in VS Code","ogDescription":"Shipping the right AI model for each task requires a lot of testing and evaluation. Get an inside look at how the VS Code and Copilot teams assess model quality, decide when to roll out updates, and balance capability with reliability.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774303544298001acRa","RegistrantKey":"1768245294609build26-1774303544298001acRa"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/HarnessBlog","backpackLink":null,"label":"Read the deep dive The Coding Harness Behind GitHub Copilot in VS Code","category":"Web link"},{"link":"https://aka.ms/VSCode/DBview","backpackLink":null,"label":"Inspect the agent yourself Open the Chat Debug View in VS Code to see the prompts, tool calls, and results behind any agent run","category":"Web link"},{"link":"https://aka.ms/VSCode/GHRepo","backpackLink":null,"label":"Explore the source The harness is open source. Dig into how context assembly, tool execution, and the agent loop actually work","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e1b2cade-ecd0-4d68-b83d-057fa499fdfd","startDateTime":"2026-06-03T00:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Milos Colic, Nate Waters","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780001077828001RhtW","langLocale":"en-US","sessionId":"1780001077828001RhtW","sessionInstanceId":"1780001077828001RhtW","sessionCode":"LIVE148","sessionCodeNormalized":"EN-US-LIVE148","title":"How Xoople Scales Python for AI using Anyscale on Azure","sortTitle":"how xoople scales python for ai using anyscale on azure","description":"Python is the backbone of modern AI workflows, but scaling Python across data processing, training, and inference introduces real distributed systems challenges. In this customer conversation, Xoople’s VP of Engineering shares how their team moved from early distributed Python approaches to running production AI workloads with Ray on Azure.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video begins with host Nate Waters from Azure introducing himself and welcoming guest Milos Kolek from Zoople (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>). They discuss that the focus of the conversation will be on scaling Python AI workloads on Azure (<a class=\"medius-aisummary\" data-starttime=\"00:00:10\">00:00:10</a>). Milos explains that Zoople—more accurately pronounced “Zupal”—is a Spanish-based global startup building an “Earth system of record,” a foundational Earth data infrastructure designed for the AI era (<a class=\"medius-aisummary\" data-starttime=\"00:00:31\">00:00:31</a>). The goal is to empower businesses to make better, high-stakes decisions using AI-driven reasoning based on real-world assets and environmental data.</p>\n\n<p><strong>Company Background and Use Cases:</strong> Milos details that Zupal’s customers span diverse industries such as agriculture, insurance, and critical infrastructure, all of which rely on physical assets exposed to environmental risk (<a class=\"medius-aisummary\" data-starttime=\"00:01:29\">00:01:29</a>). Their AI platform enables these organizations to interpret satellite imagery and environmental data for decision-making. He gives an example of a large-scale infrastructure project—like a railway network—where satellite-observed changes such as vegetation overgrowth or unauthorized building can impact outcomes (<a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>). Zupal helps businesses track and respond to these changes dynamically, combining logical and physical data to improve operational efficiency in a constantly evolving world.</p>\n\n<p><strong>Technical Implementation with AnyScale on Azure:</strong> Nate transitions to discuss Microsoft’s new announcement—“AnyScale on Azure” (<a class=\"medius-aisummary\" data-starttime=\"00:02:56\">00:02:56</a>). Built on Azure Kubernetes Service, it integrates the open-source distributed compute framework Ray into Azure to simplify scaling AI workloads. Milos explains how Zupal uses this to handle data ingestion, training, and inference (<a class=\"medius-aisummary\" data-starttime=\"00:03:18\">00:03:18</a>). Their system leverages hybrid compute—combining CPUs for data preparation and GPUs for large model computation, including processing multimodal satellite imagery reaching trillions of pixels. A key example is using IBM and NASA’s foundational model, “Teramind,” to classify land use via embeddings and clustering (<a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>).</p>\n\n<p><strong>Engineering Strategy and Performance:</strong> Discussing build-versus-buy tradeoffs, Milos emphasizes Zupal’s choice to adopt managed services like AnyScale on Azure instead of building distributed systems infrastructure from scratch (<a class=\"medius-aisummary\" data-starttime=\"00:05:31\">00:05:31</a>). This decision allows their engineers to focus on solving domain-specific problems rather than managing clusters. Milos highlights strong collaboration with forward-deployed engineers from AnyScale and Azure, which helped Zupal achieve GPU utilization rates as high as 80–90% efficiency (<a class=\"medius-aisummary\" data-starttime=\"00:07:05\">00:07:05</a>). The tight integration between CPU data handling and GPU processing in Azure pipelines optimized their end-to-end performance, exemplifying successful partnership-driven development.</p>\n\n<p><strong>Best Practices and Future Outlook:</strong> Turning to advice for other developers, Milos advises approaching scalability by breaking problems into “smallest solvable units” that can be parallelized—a key to performance and roadmap efficiency (<a class=\"medius-aisummary\" data-starttime=\"00:08:06\">00:08:06</a>). He emphasizes starting with clear problem definitions rather than focusing prematurely on models or metrics, echoing the philosophy “fall in love with the problem, not the solution” (<a class=\"medius-aisummary\" data-starttime=\"00:12:38\">00:12:38</a>). He envisions the future of AI and engineering evolving toward “hypervelocity,” where generative AI and agent-based systems augment human engineers. Zupal already employs AI-powered tools like GitHub Copilot on Azure to enhance productivity, allowing engineers and product teams to collaborate more closely and iterate faster (<a class=\"medius-aisummary\" data-starttime=\"00:11:06\">00:11:06</a>).</p>\n\n<p><strong>Conclusion and Partnership Reflections:</strong> In closing, Milos praises the deep collaboration between Microsoft and AnyScale, describing Zupal as a “partner-first” company that thrives when technology vendors strengthen their integration (<a class=\"medius-aisummary\" data-starttime=\"00:13:26\">00:13:26</a>). The conversation concludes with Nate thanking Milos for joining Microsoft Build and encouraging viewers to explore AnyScale on Azure, which is now available in public preview (<a class=\"medius-aisummary\" data-starttime=\"00:14:06\">00:14:06</a>). The interview wraps up on a positive note celebrating the fusion of open-source toolchains, managed cloud infrastructure, and applied AI innovation for scaling global data workloads efficiently.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"00:10 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/20919845-b4f7-4ac7-ad3f-4a954711c3b5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/20919845-b4f7-4ac7-ad3f-4a954711c3b5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/20919845-b4f7-4ac7-ad3f-4a954711c3b5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/20919845-b4f7-4ac7-ad3f-4a954711c3b5","speakerIds":["1768245294609build26-1778401308560001IGDK","1768245294609build26-1778017586274001FkM6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/20919845-b4f7-4ac7-ad3f-4a954711c3b5","ogTitle":"How Xoople Scales Python for AI using Anyscale on Azure","ogDescription":"Python is the backbone of modern AI workflows, but scaling Python across data processing, training, and inference introduces real distributed systems challenges. In this customer conversation, Xoople’s VP of Engineering shares how their team moved from early distributed Python approaches to running production AI workloads with Ray on Azure.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778401308560001IGDK","RegistrantKey":"1768245294609build26-1778401308560001IGDK"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778017586274001FkM6","RegistrantKey":"1768245294609build26-1778017586274001FkM6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"bb74bd60-2522-4169-a2fd-125f3c7e2112","startDateTime":"2026-06-02T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Patrick Brown, Rohan Ganesh, Surya Kukkapalli, Troy McBride, Vivaan Sawant, Advika Vuppala, Hans Yang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778699194568001Vqb4","langLocale":"en-US","sessionId":"1778699194568001Vqb4","sessionInstanceId":"1778699194568001Vqb4","sessionCode":"LIVE199","sessionCodeNormalized":"EN-US-LIVE199","title":"Imagine Cup World Championship","sortTitle":"imagine cup world championship","description":"The final moment is here. Three startups. One stage. This year’s finalists take the stage pitching their solutions as they compete to become the 2026 Imagine Cup World Champion, win $100,000 USD, and a mentorship session with Microsoft Chairman and CEO, Satya Nadella. This is where ideas meet execution and where student founders show what it looks like to build with scale in mind.","aiDescription":"<p><strong>Opening and Founders' Reflections:</strong> The video begins with brief introductions from previous Imagine Cup participants sharing their work and experiences. One founder explains that their company, Caduce, is focused on improving insurance reimbursement documentation for healthcare providers (<a class=\"medius-aisummary\" data-starttime=\"00:00:33\">00:00:33</a>), while another describes Prada Health, a telehealth addiction clinic leveraging AI to improve quality of care (<a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>). They credit mentorship sessions, especially one with Satya Nadella, for helping them refine their business approach. The reflections emphasize how the competition accelerated their understanding of scaling systems and deploying their ideas. One founder expresses hope that new teams will gain just as much from participating in the Imagine Cup and Microsoft Build experiences (<a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a>).</p>\n\n<p><strong>Introduction to Imagine Cup Finals:</strong> Host Donna Sarkar takes the stage to welcome the audience to Microsoft Build (<a class=\"medius-aisummary\" data-starttime=\"00:01:31\">00:01:31</a>). She celebrates Imagine Cup’s 24-year history of supporting global student founders by connecting them with mentorship, startup ecosystems, and advanced technology resources (<a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>). Donna highlights this year’s collaboration with Microsoft for Startups, GitHub Education, and Replit, which helped participants refine and ship their solutions intentionally. From over a thousand submissions, three startups—CopyFlag, Spoilsafe, and Revora Health—rose to the top to compete in the World Championship round (<a class=\"medius-aisummary\" data-starttime=\"00:03:06\">00:03:06</a>). Donna introduces the prestigious judging panel and announces that the winning team will receive $100,000 and a mentorship session with Satya Nadella (<a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>).</p>\n\n<p><strong>Finalist Introductions and Projects Overview:</strong> The video transitions to short documentary-style vignettes introducing each finalist and their projects. Patrick, founder of CopyFlag, recounts losing significant income due to copyright infringement, which inspired him to create an AI-powered protection platform for independent creators (<a class=\"medius-aisummary\" data-starttime=\"00:03:16\">00:03:16</a>). Spoilsafe’s team discusses global food waste inefficiencies and presents their sensor-based AI system designed to predict spoilage across supply chains, improving distribution and sustainability outcomes (<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>). The founders of Revora Health explain their platform for connecting patients with physical therapists quickly and affordably, leveraging AI to personalize care and reduce provider fatigue (<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>).</p>\n\n<p><strong>Live Demonstrations:</strong> During the live portion of the event, each team presents their demo to the audience and judges. CopyFlag showcases how users can upload content for fingerprinting and AI-driven search to detect and remove stolen works (<a class=\"medius-aisummary\" data-starttime=\"00:06:45\">00:06:45</a>). It features real-time alerts, predictive models like “Horizon,” and pattern mapping for repeat offenders, turning reactive copyright responses into predictive defense mechanisms. Spoilsafe’s founders demonstrate how their sensors gauge environmental conditions like temperature and humidity to predict spoilage for specific foods, providing actionable decisions that help warehouse teams prevent waste (<a class=\"medius-aisummary\" data-starttime=\"00:08:21\">00:08:21</a>). Revora Health’s team illustrates an intuitive interface where users interact with an AI assistant named Clara to assess injuries and receive dynamically updated therapy plans shared with physical therapists in real time (<a class=\"medius-aisummary\" data-starttime=\"00:09:43\">00:09:43</a>).</p>\n\n<p><strong>Founder Insights and Mentorship Session:</strong> Hans Yang, Vice President at Microsoft for Startups, hosts a discussion to uncover lessons from the finalists (<a class=\"medius-aisummary\" data-starttime=\"00:11:23\">00:11:23</a>). Surya from Revora shares that trust in healthcare is built from the ground up and emphasizes co-designing with therapists to create validated, practical solutions (<a class=\"medius-aisummary\" data-starttime=\"00:12:11\">00:12:11</a>). Rohan from Spoilsafe reflects on a pivotal customer insight—“We don’t need more data, we need actions”—that shifted their focus from selling visibility to selling confidence (<a class=\"medius-aisummary\" data-starttime=\"00:13:07\">00:13:07</a>). Patrick from CopyFlag discusses transforming personal loss into an innovative venture that helps creators visualize and combat unseen forms of digital theft (<a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>). Hans praises their clarity, drive, and problem-solving acumen, underscoring their symbolic role as model founders.</p>\n\n<p><strong>Announcement and Conclusion:</strong> As the excitement peaks, Donna and Hans congratulate the teams and acknowledge their mentors for helping shape their journeys (<a class=\"medius-aisummary\" data-starttime=\"00:15:03\">00:15:03</a>). Hans reveals a surprise announcement: Microsoft has increased the grand prize from $100,000 to $150,000 (<a class=\"medius-aisummary\" data-starttime=\"00:16:07\">00:16:07</a>). Finally, amid cheers from the audience, CopyFlag is declared the 2026 Imagine Cup World Champion (<a class=\"medius-aisummary\" data-starttime=\"00:16:19\">00:16:19</a>). The program concludes by celebrating innovation, resilience, and the spirit of global entrepreneurship that Imagine Cup continues to foster.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"16:15 - 16:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T16:30:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e4271f82-0b86-40ab-a575-c7596029ed2e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e4271f82-0b86-40ab-a575-c7596029ed2e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e4271f82-0b86-40ab-a575-c7596029ed2e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e4271f82-0b86-40ab-a575-c7596029ed2e","speakerIds":["1768245294609build26-1776076731812001gK5F","1768245294609build26-1775792452897001CnyX","1768245294609build26-1775834985858001az7i","1768245294609build26-1775882816983001k1U7","1768245294609build26-1776040528117001fUtV","1768245294609build26-1776049446660001RE11","1768245294609build26-1774465760748001jIh3"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.233Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e4271f82-0b86-40ab-a575-c7596029ed2e","ogTitle":"Imagine Cup World Championship","ogDescription":"The final moment is here. Three startups. One stage. This year’s finalists take the stage pitching their solutions as they compete to become the 2026 Imagine Cup World Champion, win $100,000 USD, and a mentorship session with Microsoft Chairman and CEO, Satya Nadella. This is where ideas meet execution and where student founders show what it looks like to build with scale in mind.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776076731812001gK5F","RegistrantKey":"1768245294609build26-1776076731812001gK5F"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775834985858001az7i","RegistrantKey":"1768245294609build26-1775834985858001az7i"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776049446660001RE11","RegistrantKey":"1768245294609build26-1776049446660001RE11"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775882816983001k1U7","RegistrantKey":"1768245294609build26-1775882816983001k1U7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775792452897001CnyX","RegistrantKey":"1768245294609build26-1775792452897001CnyX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776040528117001fUtV","RegistrantKey":"1768245294609build26-1776040528117001fUtV"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774465760748001jIh3","RegistrantKey":"1768245294609build26-1774465760748001jIh3"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Azure Copilot","logicalValue":"Azure Copilot"}]},{"@search.score":1.0,"documentId":"84a4cf1f-c8db-4116-b64a-088499129a0a","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Bryan Hiestand","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213122343001nkzg","langLocale":"en-US","sessionId":"1778213122343001nkzg","sessionInstanceId":"1778213122343001nkzg","sessionCode":"LIVESP122","sessionCodeNormalized":"EN-US-LIVESP122","title":"Intel quick take: Live from Microsoft Build","sortTitle":"intel quick take: live from microsoft build","description":"Get a sneak peek of the latest from Intel featuring Microsoft’s Karuana Gatimu and Bryan Hiestand","aiDescription":"<p><strong>Introduction and Setting:</strong> The video opens with host Karuana Gatimu welcoming viewers back to the Show Floor at Microsoft Build (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>). She expresses excitement about being at the Intel booth and introduces Bryan Hiestand as her guest. Bryan is joining for his first Microsoft Build event, and Karuana warmly welcomes him (<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>).</p>\n\n<p><strong>Overview of the Microsoft–Intel Partnership:</strong> Bryan begins by explaining the long-standing relationship between Intel and Microsoft (<a class=\"medius-aisummary\" data-starttime=\"00:00:26\">00:00:26</a>). He emphasizes that the partnership has lasted decades and that Intel has supported the developer community for almost 30 years (<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>). Their joint mission is to assist developers in optimizing applications across the compute continuum — ensuring workloads run on the most effective hardware, whether it is CPU, GPU, or NPU (<a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:58\">00:00:58</a>).</p>\n\n<p><strong>Empowering Developers with AI PC Technology:</strong> When Karuana asks how developers can get started, Bryan explains that Intel now has over 100 million AI PCs deployed in the market (<a class=\"medius-aisummary\" data-starttime=\"00:01:13\">00:01:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>). These systems are capable of running workloads optimized across multiple compute infrastructures. He adds that Intel offers software toolkits to help developers fine-tune their applications to perform best on the right infrastructure for their use case (<a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>).</p>\n\n<p><strong>Focus on Security and Confidential Computing:</strong> Karuana then raises the issue of data protection and security concerns (<a class=\"medius-aisummary\" data-starttime=\"00:01:34\">00:01:34</a>). Bryan responds by emphasizing the importance of security in the AI era and introduces Intel’s technology called TDX (<a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>). TDX enables confidential computing within Microsoft Azure, allowing developers to run sensitive workloads and data in protected environments, ensuring that confidentiality and integrity are maintained throughout processing (<a class=\"medius-aisummary\" data-starttime=\"00:01:53\">00:01:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:59\">00:01:59</a>).</p>\n\n<p><strong>Wrap-Up and Call to Action:</strong> The conversation concludes with Karuana encouraging viewers to explore Intel’s contributions further through the Partner Showcase page (<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a>). She invites attendees to learn more about the partnership, Intel’s developer support, and secure computing initiatives presented at Microsoft Build.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ac4e728e-13e3-415e-bae4-1e118bc96b6a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ac4e728e-13e3-415e-bae4-1e118bc96b6a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ac4e728e-13e3-415e-bae4-1e118bc96b6a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ac4e728e-13e3-415e-bae4-1e118bc96b6a","speakerIds":["1768245294609build26-1777915792311001KROB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.772Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Intel"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ac4e728e-13e3-415e-bae4-1e118bc96b6a","ogTitle":"Intel quick take: Live from Microsoft Build","ogDescription":"Get a sneak peek of the latest from Intel featuring Microsoft’s Karuana Gatimu and Bryan Hiestand","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital, Speaker In-person","SpeakerId":"1768245294609build26-1777915792311001KROB","RegistrantKey":"1768245294609build26-1777915792311001KROB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"96bb18a4-aba1-4abd-a815-8750a5e0610a","startDateTime":"2026-06-03T20:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Anthony Shaw","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780009492404001CiHJ","langLocale":"en-US","sessionId":"1780009492404001CiHJ","sessionInstanceId":"1780009492404001CiHJ","sessionCode":"LIVE165","sessionCodeNormalized":"EN-US-LIVE165","title":"Is DOOM a Tensor?","sortTitle":"is doom a tensor?","description":"Every model you use runs on tensors. But what actually is a tensor, where does DOOM fit in, and why does it matter when you are trying to get Copilot to optimize your code? Anthony Shaw breaks down how machine learning models work under the hood and what knowing that changes about how you write prompts that actually get results.","aiDescription":"<p><strong>Introduction and Setup:</strong> The session opens with friendly banter between the hosts as they reintroduce themselves at <a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>. Anthony Shaw is introduced as a technical advisor and former Python advocate at Microsoft, having contributed to Python-related work for several years. The topic quickly shifts to the intriguing and humorous premise of the segment—asking, “Is Doom a tensor?” (<a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>). Although initially unclear, the speakers promise to make sense of the question by connecting concepts from Python, machine learning, and Onnx models within the next 15–20 minutes. </p>\n\n<p><strong>Explaining Tensors and Machine Learning Models:</strong> Shaw begins defining what a tensor is around <a class=\"medius-aisummary\" data-starttime=\"00:01:23\">00:01:23</a>, describing it as data with size and shape — commonly represented as matrices. He provides an example of a simple machine learning operation that sums array items (<a class=\"medius-aisummary\" data-starttime=\"00:01:32\">00:01:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>). The discussion then introduces Netron, a tool to visually inspect and understand AI models (<a class=\"medius-aisummary\" data-starttime=\"00:02:17\">00:02:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>). They explore a specific model called “Harrier,” noting how tokens and attention masks form part of its input. Shaw explains that machine learning models consist of both weights and executable instructions — not just static data — illustrating how models execute layers to transform input sequences into tokenized outputs (<a class=\"medius-aisummary\" data-starttime=\"00:03:12\">00:03:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:21\">00:03:21</a>). </p>\n\n<p><strong>Connecting ONNX Models and Hardware Optimization:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>, conversation turns to hardware optimization and the Onnx platform’s flexibility. Shaw references working with NVIDIA on projects involving new Surface Ultra devices and notes that Onnx is Turing complete—meaning, theoretically, anything computable could be represented within it. This idea sparks the question: if Onnx supports complex logic, could it run the video game Doom? (<a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>). Burk playfully doubts this, leading Shaw to reference the famous “Potato Doom” experiment, where potatoes were used to power a device that ran the game (<a class=\"medius-aisummary\" data-starttime=\"00:05:28\">00:05:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a>). Shaw then outlines three potential methods for running Doom in an Onnx environment: compiling its source code into Onnx, creating a CPU emulator, or utilizing other creative methods. The audience votes, and option two – emulating a CPU – is chosen (<a class=\"medius-aisummary\" data-starttime=\"00:06:15\">00:06:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:28\">00:06:28</a>). </p>\n\n<p><strong>Building and Running Doom in Excel and Onnx:</strong> Shaw explains that he used the RISC-V architecture as a baseline to emulate a CPU within Onnx nodes (<a class=\"medius-aisummary\" data-starttime=\"00:06:42\">00:06:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>). He then demonstrates compiling Doom from source into machine code, and surprisingly, shows that machine code can be read and partially executed in Microsoft Excel formula logic (<a class=\"medius-aisummary\" data-starttime=\"00:08:07\">00:08:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:20\">00:08:20</a>). Shaw confirms that Excel is Turing complete and describes building a spreadsheet CPU that can interpret Doom’s code line by line — only limited by Excel’s one-million-row maximum. Despite performance challenges, this playful experiment becomes part of Microsoft Build 2026’s announcements, showing “Doom runs in Excel” in a retro AI-generated promotion. In another demo using Netron, Shaw presents the Onnx-based CPU graph with tensors acting as RAM and registers (<a class=\"medius-aisummary\" data-starttime=\"00:10:32\">00:10:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:11\">00:11:11</a>). The system indeed renders a Doom frame — taking approximately six hours per single frame, equating to a 2 kHz CPU speed (<a class=\"medius-aisummary\" data-starttime=\"00:12:03\">00:12:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:18\">00:12:18</a>). </p>\n\n<p><strong>Optimizing Performance and Lessons Learned:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:12:26\">00:12:26</a> onward, Shaw transitions the conversation toward meaningful lessons in performance tuning and AI optimization. He shares how Copilot assisted in iteration but struggled with caching and computational overhead. He emphasizes standard approaches for improving AI model speed—caching outputs, reducing work, and scaling hardware. Despite limited success (progressing from 2 kHz to 4 kHz CPU speed), he illustrates that optimization requires iterative improvement and hardware consideration (<a class=\"medius-aisummary\" data-starttime=\"00:13:04\">00:13:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:47\">00:13:47</a>). Shaw also describes how to “coach” AI agents to improve speed by using benchmarks and rubber-duck debugging within Copilot CLI. The final accelerated demo shows Doom’s playback simulated via Onnx outputs converted to GIF frames, though without sound or keyboard input (<a class=\"medius-aisummary\" data-starttime=\"00:15:01\">00:15:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:19\">00:15:19</a>). </p>\n\n<p><strong>Conclusion and Key Insights:</strong> As the talk closes around <a class=\"medius-aisummary\" data-starttime=\"00:15:54\">00:15:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:09\">00:17:09</a>, Shaw humorously confirms that “Doom is a tensor” but clarifies that while the experiment proved theoretical completeness, it is not practical or efficient. The key takeaway is that AI model nodes and instructions must be optimized per hardware configuration to maximize performance. He cautions that downloading models from repositories should include checking compatibility with the user’s CPU or GPU due to differing tensor implementations. The hosts conclude with appreciation and laughter, noting the educational depth and humor of the session. The segment ends as they thank viewers and sign off. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"20:40 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":20,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/520dcf59-8b05-4f90-aee1-42ae648865ab","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/520dcf59-8b05-4f90-aee1-42ae648865ab","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/520dcf59-8b05-4f90-aee1-42ae648865ab","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/520dcf59-8b05-4f90-aee1-42ae648865ab","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1770049778212001cejZ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/520dcf59-8b05-4f90-aee1-42ae648865ab","ogTitle":"Is DOOM a Tensor?","ogDescription":"Every model you use runs on tensors. But what actually is a tensor, where does DOOM fit in, and why does it matter when you are trying to get Copilot to optimize your code? Anthony Shaw breaks down how machine learning models work under the hood and what knowing that changes about how you write prompts that actually get results.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770049778212001cejZ","RegistrantKey":"1768245294609build26-1770049778212001cejZ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Build your VS Code and Copilot skills with free, hands-on courses","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"09141a2d-e6f2-451d-893c-dcf64713af54","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Stephen McCullough","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213403494001JTE2","langLocale":"en-US","sessionId":"1778213403494001JTE2","sessionInstanceId":"1778213403494001JTE2","sessionCode":"LIVESP123","sessionCodeNormalized":"EN-US-LIVESP123","title":"NVIDIA quick take: Live from Microsoft Build","sortTitle":"nvidia quick take: live from microsoft build","description":"Get a sneak peek of the latest from NVIDIA featuring Microsoft’s Karuana Gatimu and Stephen McCullough","aiDescription":"<p><strong>Introduction and Welcome:</strong> The video opens with the host welcoming viewers to the Microsoft Build event <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>. The host enthusiastically announces a special guest from NVIDIA, Steven, who will be sharing details about innovative projects his team has been developing. After a brief greeting exchange between Caruana and Steven <a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>, the host expresses excitement about learning what Steven has been building in partnership with Microsoft.</p>\n\n<p><strong>Introduction to the Collaboration:</strong> Steven begins by discussing the collaborative relationship between Microsoft and NVIDIA <a class=\"medius-aisummary\" data-starttime=\"00:00:30\">00:00:30</a>. He mentions that both companies have been building several exciting technologies together, many of which are being showcased at the Build event. One special focus of his work is the integration involving hosted agents, Nemotron, and Hermes Agent <a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>. Steven notes that a deeper look into this project will be featured in a talk by Joey Conway, encouraging the audience to attend that session <a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>.</p>\n\n<p><strong>Highlighting the Technological Innovations:</strong> The host responds with enthusiasm about the upcoming demonstration, emphasizing the value of seeing technology in action <a class=\"medius-aisummary\" data-starttime=\"00:00:57\">00:00:57</a>. Caruana stresses that tangible demonstrations help audiences better understand and appreciate the practical impact of new solutions. She prompts Steven to share what makes this new implementation important and why it matters for customers working with these integrations <a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a>.</p>\n\n<p><strong>Technical Deep Dive and Benefits:</strong> Steven elaborates that the new setup solves many challenges customers face when managing agentic deployments <a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>. The first key benefit he mentions is improved control over agentic environments. Hosted agents provide strong opportunities for customization, while Nemotron contributes advanced intelligence and open-source flexibility <a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>. The Hermes Agent then ties everything together, orchestrating models efficiently to maximize their performance and value <a class=\"medius-aisummary\" data-starttime=\"00:01:37\">00:01:37</a>. Steven describes this combination as a \"one-two-three punch\" of innovation.</p>\n\n<p><strong>Showcase and Closing Remarks:</strong> Wrapping up, Steven expresses excitement about showcasing these technologies at the NVIDIA exhibit within the Build conference <a class=\"medius-aisummary\" data-starttime=\"00:01:47\">00:01:47</a>. The host encourages viewers to visit the NVIDIA showcase page to explore these projects firsthand <a class=\"medius-aisummary\" data-starttime=\"00:01:55\">00:01:55</a>. As the conversation concludes, Caruana thanks Steven for joining the session and expresses enthusiasm about reconnecting soon, while Steven reciprocates his appreciation before the session ends <a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c4d84be5-ddcd-494b-9489-a35eb76b26ed","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c4d84be5-ddcd-494b-9489-a35eb76b26ed","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c4d84be5-ddcd-494b-9489-a35eb76b26ed","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c4d84be5-ddcd-494b-9489-a35eb76b26ed","speakerIds":["1768245294609build26-1769795191086001hVae"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.772Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c4d84be5-ddcd-494b-9489-a35eb76b26ed","ogTitle":"NVIDIA quick take: Live from Microsoft Build","ogDescription":"Get a sneak peek of the latest from NVIDIA featuring Microsoft’s Karuana Gatimu and Stephen McCullough","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769795191086001hVae","RegistrantKey":"1768245294609build26-1769795191086001hVae"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"bad96cbc-97a1-4889-95a3-6e6756e31ba1","startDateTime":"2026-06-02T23:05:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Max Kirby","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213548705001EIji","langLocale":"en-US","sessionId":"1778213548705001EIji","sessionInstanceId":"1778213548705001EIji","sessionCode":"LIVESP129","sessionCodeNormalized":"EN-US-LIVESP129","title":"Organizational intelligence in action — with Anthropic","sortTitle":"organizational intelligence in action — with anthropic","description":"Join Karuana Gatimu in conversation with Anthropic’s Max Kirby to explore organizational intelligence in action. See how AI analyzes real usage to surface feature gaps and recommend what to build next, helping developers prioritize with confidence and deliver higher-impact solutions.","aiDescription":"<p><strong>Opening and Introductions:</strong> The video begins with Kirwana welcoming viewers and introducing her guest Max from Anthropic <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>. She offers greetings to both the live and digital audiences, expressing excitement for the session. Max introduces himself as someone who works at the intersection of research breakthroughs and enterprise applications <a class=\"medius-aisummary\" data-starttime=\"00:00:18\">00:00:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:26\">00:00:26</a>, discussing how he collaborates with strategic customers to determine priorities for innovation. The two celebrate the partnership between Microsoft and Anthropic, noting the buzz and energy in San Francisco during the Microsoft Build event <a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>.</p>\n\n<p><strong>Partnership Deep Dive:</strong> Kirwana asks Max to elaborate on the collaboration between Microsoft and Anthropic <a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>. Max explains how they integrate at various layers of enterprise solutions, from infrastructure to intelligence, security, and governance <a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>. He emphasizes the flexibility that Claude provides, making sure customers can deploy AI wherever needed. The partnership has been successful so far, setting the stage for future innovations. This section transitions into discussions about Microsoft’s new “IQ” announcements—Work IQ, Foundry IQ, Fabric IQ, and Web IQ—which enable reasoning across multiple domains <a class=\"medius-aisummary\" data-starttime=\"00:02:10\">00:02:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:21\">00:02:21</a>.</p>\n\n<p><strong>Concept of Legibility and Transformation:</strong> Max introduces the idea of “legibility,” describing it as the AI’s ability to understand what’s happening inside a business <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:06\">00:03:06</a>. He explains that AI can’t improve processes it can’t see, and legibility allows enterprises to make operations more transparent and interpretable. Using examples, Max highlights how many companies don’t have a full view of their own standard procedures or implicit know-how, stored within employees rather than documentation <a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>. By improving AI visibility, leaders can make better-informed decisions, bridging the gap between executive strategy and on-the-ground operations. This visibility then guides what technology should be built—a reversal from traditional development sequences that focused on building before observing <a class=\"medius-aisummary\" data-starttime=\"00:04:32\">00:04:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>.</p>\n\n<p><strong>Demo and Technical Insights:</strong> The demo segment follows <a class=\"medius-aisummary\" data-starttime=\"00:05:07\">00:05:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:10\">00:05:10</a>, showcasing experiments around legibility where Claude identifies patterns from transcripts and user actions. Kirwana praises this new method of deriving actionable insights for development teams. Max explains two major focal areas for Claude—understanding complex codebases and showing empathy toward human users <a class=\"medius-aisummary\" data-starttime=\"00:06:04\">00:06:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>. Together, these make possible a new paradigm: “observe first, build second.” Instead of guessing what users need, developers can watch interactions to infer the next valuable solutions. Both express enthusiasm about how this insight-driven approach transforms productivity and user-value generation <a class=\"medius-aisummary\" data-starttime=\"00:07:02\">00:07:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:34\">00:07:34</a>.</p>\n\n<p><strong>Business Impact and Leadership Evolution:</strong> Kirwana and Max discuss how AI can drive innovation rather than just efficiency, enabling businesses to grow and explore new possibilities <a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:05\">00:08:05</a>. They highlight how high intelligence removes blockers and supports problem-solving, even to the extent that “asking Claude what to do next” becomes a viable business strategy <a class=\"medius-aisummary\" data-starttime=\"00:08:36\">00:08:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:43\">00:08:43</a>. Kirwana emphasizes leadership traits required in the AI era—courage and release of control—sharing her own experience of learning to collaborate with AI tools like Copilot. Max agrees, noting that vulnerability and openness from senior leaders in using AI lead to greater organizational adoption. Demonstrating real-world AI use within teams helps inspire trust and drives higher transformation rates across enterprises <a class=\"medius-aisummary\" data-starttime=\"00:10:02\">00:10:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:42\">00:10:42</a>.</p>\n\n<p><strong>Conclusion and Forward Outlook:</strong> As the session wraps up <a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:27\">00:11:27</a>, Max and Kirwana reflect on the compatibility and future potential of their partnership. They discuss how earlier AI adoption involved manual steps like copy-paste, evolving toward tool-based automation and now toward intelligent context sharing with technologies like MCP <a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:16\">00:12:16</a>. Both agree that their joint “North Star” should be enabling data-driven root cause analysis and improvement. Kirwana closes by encouraging viewers to explore deeper demos on the event showcase page and to engage within the developer community <a class=\"medius-aisummary\" data-starttime=\"00:12:28\">00:12:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:03\">00:13:03</a>. She ends with the belief that AI development is a “team sport,” emphasizing collaboration and continuous learning among all builders and enterprise leaders as the key to maximizing the benefits of artificial intelligence.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"23:05 - 23:20","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:20:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8a1f6f67-1328-4be4-8a17-c0fff1c551ce","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8a1f6f67-1328-4be4-8a17-c0fff1c551ce","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8a1f6f67-1328-4be4-8a17-c0fff1c551ce","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8a1f6f67-1328-4be4-8a17-c0fff1c551ce","speakerIds":["1768245294609build26-1779375673264001Y4xi"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.773Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":["Anthropic","PBC"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8a1f6f67-1328-4be4-8a17-c0fff1c551ce","ogTitle":"Organizational intelligence in action — with Anthropic","ogDescription":"Join Karuana Gatimu in conversation with Anthropic’s Max Kirby to explore organizational intelligence in action. See how AI analyzes real usage to surface feature gaps and recommend what to build next, helping developers prioritize with confidence and deliver higher-impact solutions.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779375673264001Y4xi","RegistrantKey":"1768245294609build26-1779375673264001Y4xi"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"de8ffdb4-af63-45d1-a204-b700a144f725","startDateTime":"2026-06-02T22:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chip Huyen, John Maeda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779989545323001gkm3","langLocale":"en-US","sessionId":"1779989545323001gkm3","sessionInstanceId":"1779989545323001gkm3","sessionCode":"LIVE141","sessionCodeNormalized":"EN-US-LIVE141","title":"Programming robots","sortTitle":"programming robots","description":"AI is taking over the digital world, and it's exciting to see how AI can be used in the physical world. Once we have a robot, we need to figure out how to program it to do what we want. This demo shows how we can interface with multiple robots using the same API.","aiDescription":"<p><strong>Introduction and the ChatGPT Moment:</strong> The discussion begins with the host welcoming the audience and introducing guest Chip Yan, who is known for his expertise in AI engineering <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:09\">00:00:09</a>. Chip recounts his personal reaction to the launch of ChatGPT, describing an intense two-month existential crisis triggered by the realization that AI could now perform both of his core skills—writing and coding—with alarming competence <a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>. This period of uncertainty led him to evaluate AI’s broader implications and distill his understanding into his book on AI engineering <a class=\"medius-aisummary\" data-starttime=\"00:01:13\">00:01:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>. He emphasizes how even experts underestimated the exponential usefulness of small performance improvements in AI systems.</p>\n\n<p><strong>Human Perception and Limits of Digital Intelligence:</strong> Chip explores why AI’s accelerating fluency caught many by surprise, arguing that humans are poor at assessing usefulness relative to incremental accuracy gains <a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>. He further contrasts the relative simplicity of modeling the human mind versus the human body, noting that while AI can write code or essays effortlessly, it still struggles with seemingly simple physical actions, such as teaching a humanoid robot to sit gracefully <a class=\"medius-aisummary\" data-starttime=\"00:02:20\">00:02:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:12\">00:03:12</a>. Unlike the digital world, where structured documentation exists (e.g., for APIs), the physical world lacks predictable parameters—objects can break, children can be harmed, and tactile feedback must be learned experientially <a class=\"medius-aisummary\" data-starttime=\"00:03:36\">00:03:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>.</p>\n\n<p><strong>Robotics Progress and Challenges:</strong> Transitioning from theory to application, Chip discusses the complexities of building robots that reason and move effectively <a class=\"medius-aisummary\" data-starttime=\"00:04:47\">00:04:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:59\">00:06:59</a>. He cites Unitary, a profitable robotics firm pursuing an IPO, whose CEO frames robotic intelligence as a combination of reasoning (planning tasks such as fetching water) and motion (executing precise actions). Modern AI excels at step-by-step reasoning, but still struggles with physical feedback—knowing how much pressure to apply without causing damage. Chip highlights research into “world modeling,” where AI learns environmental rules and dynamics, backed by major investments from various startups <a class=\"medius-aisummary\" data-starttime=\"00:06:39\">00:06:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:56\">00:06:56</a>. For motion, many current robot demonstrations are preprogrammed, and the ultimate challenge remains enabling robots to generate adaptive, natural movements autonomously <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>.</p>\n\n<p><strong>Integrating Vision, Language, and Motion Models:</strong> When asked about connecting vision and motion, Chip describes two main research directions: vision-language-action (VLA) models and world-model-based systems <a class=\"medius-aisummary\" data-starttime=\"00:07:49\">00:07:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>. VLA approaches ingest visual data and language instructions to generate physical actions, but collecting action data is expensive and often incompatible across robot designs. In contrast, world modeling focuses on encoding environmental properties separately. Researchers are divided on which approach will prevail, with competing camps labeling the other as misguided. Chip concludes that the field remains in flux, and success will be clear only once a model truly works in practice <a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>.</p>\n\n<p><strong>Guidance for Developers and Industry Safety Lessons:</strong> Moving from research to application, Chip advises developers to stay focused instead of chasing every AI advance or funding announcement <a class=\"medius-aisummary\" data-starttime=\"00:09:27\">00:09:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:37\">00:10:37</a>. He suggests filtering innovations by relevance to personal or practical problems, adopting a “blinder” mindset to maintain direction. The conversation then turns to robot safety, referencing Boston Dynamics’ Spot—a dog-like robot safer than humanoids <a class=\"medius-aisummary\" data-starttime=\"00:11:04\">00:11:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:19\">00:11:19</a>. Chip explains that humanoids present hazards—an 80‑pound falling metal body is dangerous—so safety design must include low grip strength, distance buffers, and robust battery management to prevent falls during power failures <a class=\"medius-aisummary\" data-starttime=\"00:11:39\">00:11:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:50\">00:12:50</a>.</p>\n\n<p><strong>Future Outlook on Robotics:</strong> In conclusion, Chip outlines two emerging schools in robotics development <a class=\"medius-aisummary\" data-starttime=\"00:13:16\">00:13:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:09\">00:14:09</a>. One focuses on general-purpose robots capable of performing multiple tasks like AI models do with language, though such projects are capital-intensive and data-hungry. The other emphasizes specialized robots built for narrow functions like folding laundry, cooking, or delivery. While Chip is uncertain when general-purpose robots will mature, he expresses optimism about the sector’s trajectory and closes by thanking the audience, encouraging continued innovation and thoughtful focus within AI and robotics <a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:11\">00:14:11</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"22:50 - 23:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0a938684-4ae8-403c-9495-2138587205d5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0a938684-4ae8-403c-9495-2138587205d5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0a938684-4ae8-403c-9495-2138587205d5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0a938684-4ae8-403c-9495-2138587205d5","speakerIds":["1768245294609build26-1775514714965001BQ8v","1768245294609build26-1778108772492001j4Zf"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.459Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0a938684-4ae8-403c-9495-2138587205d5","ogTitle":"Programming robots","ogDescription":"AI is taking over the digital world, and it's exciting to see how AI can be used in the physical world. Once we have a robot, we need to figure out how to program it to do what we want. This demo shows how we can interface with multiple robots using the same API.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775514714965001BQ8v","RegistrantKey":"1768245294609build26-1775514714965001BQ8v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778108772492001j4Zf","RegistrantKey":"1768245294609build26-1778108772492001j4Zf"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e69bf430-2080-4e68-8c0d-b29152def42b","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ana Schafer, Morris Novello","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778212929318001f580","langLocale":"en-US","sessionId":"1778212929318001f580","sessionInstanceId":"1778212929318001f580","sessionCode":"LIVESP126","sessionCodeNormalized":"EN-US-LIVESP126","title":"Qualcomm quick take: Live from Microsoft Build","sortTitle":"qualcomm quick take: live from microsoft build","description":"Get a sneak peek of the latest from Qualcomm featuring Microsoft’s Karuana Gatimu and Ana Schafer","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/120e8a00-0e80-4055-8bbc-1a65857c9963","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/120e8a00-0e80-4055-8bbc-1a65857c9963","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/120e8a00-0e80-4055-8bbc-1a65857c9963","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778212929318001f580_20260601103704296.avif","speakerIds":["1768245294609build26-1777661811386001ODeJ","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.772Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778212929318001f580_20260601103704296.avif","ogTitle":"Qualcomm quick take: Live from Microsoft Build","ogDescription":"Get a sneak peek of the latest from Qualcomm featuring Microsoft’s Karuana Gatimu and Ana Schafer","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777661811386001ODeJ","RegistrantKey":"1768245294609build26-1777661811386001ODeJ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"701c0b4c-1bc9-442e-98dc-6f84eaa723dd","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Hanselman, Mark Russinovich","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776792281916001rWbZ","langLocale":"en-US","sessionId":"1776792281916001rWbZ","sessionInstanceId":"1776792281916001rWbZ","sessionCode":"LIVE101","sessionCodeNormalized":"EN-US-LIVE101","title":"Scott and Mark learn to Vibe Check","sortTitle":"scott and mark learn to vibe check","description":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","aiDescription":"<p><strong>Opening and Concept Introduction:</strong> The program begins with Scott Hanselman introducing the segment “Scott and Mark Learn to Vibe Check,” joined by Mark Russinovich, who is intentionally kept in the dark about the day’s agenda <a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a>. The hosts explain their plan to humorously assess community-created AI projects to determine whether they represent “AI slop,” creative “vibes,” or genuine “AI-augmented software engineering.” They liken the format to a “Fool Us”-style session, where developers present applications and explain their implementation. The spirit is playful but aimed at showcasing how tools such as GitHub Copilot and similar technologies influence software craftsmanship. Before diving into the demonstrations, the hosts recognize sponsors for enabling such inventive programming <a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>.</p>\n\n<p><strong>Early Sponsor Highlights and AMD Partnership:</strong> The broadcast transitions briefly to the Microsoft Build show floor in San Francisco at the AMD booth <a class=\"medius-aisummary\" data-starttime=\"00:02:28\">00:02:28</a>. Scott interviews Adrian from AMD about a longstanding partnership with Microsoft. The discussion centers on fostering a developer ecosystem that enables rapid experimentation through open tooling and infrastructure. With the cost of exploration approaching zero, developers can prototype AI solutions more freely than before. Adrian highlights how this democratization of experimentation empowers creators but also adds complexity through the number of available models, frameworks, and compute configurations. Viewers are encouraged to explore related resources on AMD’s sponsor page before the show cuts back to the main vibe check segment <a class=\"medius-aisummary\" data-starttime=\"00:04:24\">00:04:24</a>.</p>\n\n<p><strong>First Vibe Check – Steve Sanderson’s “Vibe OS”:</strong> Returning to the main stage, Scott and Mark welcome software engineer Steve Sanderson, known for projects like Knockout.js and Blazor <a class=\"medius-aisummary\" data-starttime=\"00:04:49\">00:04:49</a>. Sanderson unveils “Vibe OS,” described as the world’s first “hallucinated operating system.” Demonstrating live, he boots a lightweight virtual disk file and launches familiar applications such as Notepad and Calculator. Remarkably, these interfaces function without any underlying application code—each UI element is dynamically generated by an AI model through the Copilot SDK and HTML-based rendering. Sanderson explains how user interactions are handled via fine-grained diffs fed back to the model. The hosts respond with amazement, calling it intentional slop elevated by engineering expertise. Sanderson’s build process took approximately six hours, highlighting how AI can generate novel experiential software when guided precisely <a class=\"medius-aisummary\" data-starttime=\"00:15:49\">00:15:49</a>.</p>\n\n<p><strong>Second Vibe Check – Cassidy Williams’ CSS “Database”: </strong> The next guest, Cassidy Williams, presents a project that humorously abuses browser technology by constructing a mock SQL database entirely in CSS <a class=\"medius-aisummary\" data-starttime=\"00:20:04\">00:20:04</a>. Through animated queries and “tables,” she reveals that what seems like traditional JavaScript-driven logic is entirely composed of CSS selectors, variables, and sibling indices. The hosts quiz her about the impractical yet creative effort, probing her methods and her balance between manual coding and Copilot's suggestions. Williams explains she wrote approximately 20% of the code, steering the AI to correct layout and structural issues. Despite its absurd premise, the result pushes boundaries of what “vibe coding” can mean—purposefully cursed yet technically insightful. Her experiment takes just under two hours and wins her praise for embracing both creativity and discipline in AI-assisted authoring <a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>.</p>\n\n<p><strong>Third and Fourth Vibe Checks – Swix’s AIE Bot and Simon Willison’s Data Set Agent:</strong> Later, developer and AI researcher “Swix” demonstrates an “AI Event Scheduler” for managing multi-track conferences <a class=\"medius-aisummary\" data-starttime=\"00:33:41\">00:33:41</a>. Built on Cloudflare infrastructure with an LLM-driven assistant integrated into chat, email, and web interfaces, the system automatically categorizes and arranges speaking sessions while offering human override and rollback logs. Over a week and a half, Swix developed the fully “vibe-coded” app using multiple small models rather than frontier ones, emphasizing rapid, autonomous development for real-world workflows. Finally, Simon Willison showcases his “Dataset Agent,” a Python-based plugin system leveraging locally or cloud-hosted LLMs to query and visualize his 24-year blog database <a class=\"medius-aisummary\" data-starttime=\"00:48:37\">00:48:37</a>. He describes how agents craft SQL queries, generate charts, and even run sandboxed Python securely. Willison outlines his methodology for safe AI-assisted development and adversarial model testing, exemplifying how deep domain knowledge amplifies AI’s engineering potential.</p>\n\n<p><strong>Conclusion and Reflections:</strong> The program closes with booth interviews highlighting sponsor innovations from NVIDIA, Fireworks AI, Elastic, and Qualcomm <a class=\"medius-aisummary\" data-starttime=\"01:01:00\">01:01:00</a>. Returning to the main stage, Scott and Mark reflect on all four vibe check demos. They agree that while Cassidy’s CSS database borders on pure slop, Sanderson’s Vibe OS, Swix’s scheduler, and Willison’s data agent exemplify sophisticated AI-augmented software engineering. The hosts emphasize that effective “vibe coding” requires foundational expertise and careful technical judgment to turn AI spontaneity into production-level innovation. They sign off with gratitude toward guests, sponsors, and the audience, urging viewers to continue exploring responsible, creative use of AI tools in real-world software development.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/df6b109e-848b-4a7a-b791-fcd1e8029278","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/df6b109e-848b-4a7a-b791-fcd1e8029278","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/df6b109e-848b-4a7a-b791-fcd1e8029278","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/df6b109e-848b-4a7a-b791-fcd1e8029278","speakerIds":["1768245294609build26-1769008130185001WAvh","1768245294609build26-1776631798007001nGQM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.232Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/df6b109e-848b-4a7a-b791-fcd1e8029278","ogTitle":"Scott and Mark learn to Vibe Check","ogDescription":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776631798007001nGQM","RegistrantKey":"1768245294609build26-1776631798007001nGQM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e55ce9f7-cf1d-4c36-b8ef-5e0a2b05404a","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Hanselman, Cassidy Williams","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780438953171001ODkA","langLocale":"en-US","sessionId":"1780438953171001ODkA","sessionInstanceId":"1780438953171001ODkA","sessionCode":"LIVE114","sessionCodeNormalized":"EN-US-LIVE114","title":"Scott and Mark learn to Vibe Check with Cassidy Williams","sortTitle":"scott and mark learn to vibe check with cassidy williams","description":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","aiDescription":"<p><strong>Introduction and Setup:</strong> The video opens as hosts Mark and Scott return, referencing their previous showcase by Steve Sanderson, who created a “vibe-coded operating system” in just a few evenings, demonstrating the potential of rapid prototyping with modern tools <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:10\">00:00:10</a>. They segue into introducing Cassidy Williams, inviting her to share what she has built. The hosts joke that they are unsure whether Cassidy’s work will be “AI slop,” “vibes,” or “AI-augmented software engineering” <a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>.</p>\n\n<p><strong>Cassidy’s Project Reveal:</strong> Cassidy begins by explaining her fascination with databases and her idea to build one that runs entirely within a browser tab <a class=\"medius-aisummary\" data-starttime=\"00:00:31\">00:00:31</a>. She demonstrates a query shell that appears capable of running SQL-like commands such as “SELECT * FROM ideas” and modifying queries to include filters like “WHERE cursedness = catastrophic” <a class=\"medius-aisummary\" data-starttime=\"00:00:42\">00:00:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:22\">01:22:00</a>. Initially, it seems to be a legitimate in-browser SQLite interface, but Cassidy soon reveals the twist: there is no actual database or query engine—everything is powered purely by CSS <a class=\"medius-aisummary\" data-starttime=\"00:02:15\">00:02:15</a>. The logic for “querying” is handled entirely through CSS selectors, variables, and conditional styling rules, making it both absurd and ingenious <a class=\"medius-aisummary\" data-starttime=\"00:02:20\">00:02:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:49\">00:02:49</a>.</p>\n\n<p><strong>Technical Breakdown and Philosophy:</strong> Cassidy shares how she used a combination of HTML and JavaScript only to render the interface, while CSS controlled all of the logic and data display <a class=\"medius-aisummary\" data-starttime=\"00:04:24\">00:04:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:46\">00:04:46</a>. She walks through her code, demonstrating that each table element relies on CSS variables and sibling indexes for limiting results and sorting <a class=\"medius-aisummary\" data-starttime=\"00:04:51\">00:04:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:04\">00:05:04</a>. Discussing her workflow, she explains her use of GitHub Copilot in both VS Code and the Copilot app but admits she often took over parts of the process herself, fixing issues when AI produced undesired results <a class=\"medius-aisummary\" data-starttime=\"00:05:10\">00:05:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:09\">00:07:09</a>. She estimates that around 20% of the code was handwritten, including core logic parts like ordering and flexbox manipulations <a class=\"medius-aisummary\" data-starttime=\"00:05:50\">00:05:50</a>.</p>\n\n<p><strong>CSS Exploration and Human-AI Collaboration:</strong> The conversation turns playful as the hosts discuss CSS conventions such as “!important,” flexbox behavior, and cascading rules <a class=\"medius-aisummary\" data-starttime=\"00:07:21\">00:07:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>. Cassidy highlights her creative use of flexbox to sort items for ascending and descending order within the fake database, effectively flipping UI elements using only CSS <a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:27\">00:09:27</a>. She and the hosts demonstrate extending the project live, adding another query line manually to the application and noting how cursed yet functional it is <a class=\"medius-aisummary\" data-starttime=\"00:08:13\">00:08:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>. The hosts marvel at how the AI didn’t object to such eccentric engineering choices—Cassidy emphasizes that while this project is not practical, it’s a testament to blending creativity with unconventional coding.</p>\n\n<p><strong>Reflection and Conclusion:</strong> When asked about time spent, Cassidy shares it took around two hours of work spread over roughly a week after being invited on <a class=\"medius-aisummary\" data-starttime=\"00:10:02\">00:10:02</a>. The hosts debate whether this project qualifies as “AI slop” or “AI-augmented engineering,” ultimately concluding that it sits beyond mere vibes—requiring deep CSS expertise to pull off <a class=\"medius-aisummary\" data-starttime=\"00:10:46\">00:10:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:14\">00:11:14</a>. Cassidy receives a humorous award for “Best Vibes” <a class=\"medius-aisummary\" data-starttime=\"00:11:16\">00:11:16</a>, and the hosts acknowledge her open-source contribution on GitHub, reminding her to credit Scott as a contributor for his minor code input <a class=\"medius-aisummary\" data-starttime=\"00:11:32\">00:11:32</a>. The session wraps up with thanks to sponsors and an announcement for more quirky and “cursed” coding projects coming after the break <a class=\"medius-aisummary\" data-starttime=\"00:11:41\">00:11:41</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/aabe7791-a560-45b5-8c66-f6f12a88a9c1","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/aabe7791-a560-45b5-8c66-f6f12a88a9c1","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/aabe7791-a560-45b5-8c66-f6f12a88a9c1","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/aabe7791-a560-45b5-8c66-f6f12a88a9c1","speakerIds":["1768245294609build26-1769008130185001WAvh","1768245294609build26-1774284997876001pQKi"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/aabe7791-a560-45b5-8c66-f6f12a88a9c1","ogTitle":"Scott and Mark learn to Vibe Check with Cassidy Williams","ogDescription":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","deliveryTypes":[],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284997876001pQKi","RegistrantKey":"1768245294609build26-1774284997876001pQKi"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"5344fcbe-7f3e-4554-9921-ead162040d99","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Hanselman, Simon Willison","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780435242497001tQgA","langLocale":"en-US","sessionId":"1780435242497001tQgA","sessionInstanceId":"1780435242497001tQgA","sessionCode":"LIVE113","sessionCodeNormalized":"EN-US-LIVE113","title":"Scott and Mark learn to Vibe Check with Simon Willison","sortTitle":"scott and mark learn to vibe check with simon willison","description":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","aiDescription":"<p><strong>Introduction and Guest Overview:</strong> The video opens with hosts Mark and Scott returning to continue their developer-focused series, mentioning past talks from Swix, Steve Sanderson, and Cassidy Williams before introducing the legendary Simon Willison <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>. They praise Simon for his consistent presence on Hacker News and note his reputation within the programming community. Simon is invited to showcase his latest work—the culmination of years of his projects coming together. This sets the stage for a deep technical demonstration that combines his long experience with AI-powered programming tools.</p>\n\n<p><strong>Project Background and Evolution:</strong> Simon begins describing how his various projects—starting with his long-running open source data exploration tool, \"Data Set,\" developed over eight years—have converged into his new project, the \"Data Set Agent\" <a class=\"medius-aisummary\" data-starttime=\"00:00:28\">00:00:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:13\">00:01:13</a>. He explains that the agent acts as a plugin for Data Set, adding the ability to query and analyze his own blog’s data using GPT-powered language models. He showcases a demo where the agent interacts with his 24 years of blog content, running SQL-like queries to count entries, quotes, and \"blogmarks\" <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:47\">00:01:47</a>. This marks the integration of AI and traditional software tools into his workflow and demonstrates how his local environment uses copies of his blog’s database to experiment with these models offline <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>.</p>\n\n<p><strong>Plugin Architecture and Technical Demo:</strong> The discussion then moves deeper into the concept of plugins. Simon explains how his architecture in Python leverages dynamic plugin registration, enabling modular expansion across different tools <a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:49\">00:04:49</a>. He demonstrates a newly published plugin named \"Data Set Agent Micropython\" that allows Python code execution within a secure WebAssembly sandbox—essential for preventing agents from unsafe system access <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>. Simon elaborates on how his plugin system consists of hundreds of extensions, supporting various functionalities such as charting and local code execution. The hosts highlight how this modular design reflects a modern approach to AI-augmented software engineering by enabling quick and safe experimentation with new functionalities.</p>\n\n<p><strong>Vibe Coding and AI-Augmented Development Practices:</strong> The conversation shifts toward the concept of \"vibe coding\"—a creative, rapid prototyping method using AI tools to generate lightweight HTML and JavaScript prototypes <a class=\"medius-aisummary\" data-starttime=\"00:05:31\">00:05:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:04\">00:07:04</a>. Simon recalls starting this approach about two and a half years ago when AI models improved at generating interactive web tools. He now maintains over 200 mini web apps under his domain \"tools.simonwillison.net\" <a class=\"medius-aisummary\" data-starttime=\"00:06:14\">00:06:14</a>. He demonstrates an example UI prototype, explaining how coding agents like Claude and Codex assist him in generating workable software in minutes. Most of his recent code, he notes, has been written by agents rather than manually typed, underscoring a shift toward directing development through prompt engineering rather than traditional programming.</p>\n\n<p><strong>Adversarial and Secure Development Methods:</strong> Simon and the hosts discuss techniques such as adversarial vibe coding, where multiple AI models critique each other’s code for better accuracy and security <a class=\"medius-aisummary\" data-starttime=\"00:08:22\">00:08:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:17\">00:09:17</a>. Simon shares that he often uses GPT 5.5 as a \"security blanket\" to review outputs from other models like Opus and Claude, detecting vulnerabilities and errors. They also talk about his method of monitoring code generation depending on project stakes—for critical components, he closely audits AI output, while low-risk prototypes receive minimal oversight <a class=\"medius-aisummary\" data-starttime=\"00:07:57\">00:07:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:15\">00:08:15</a>. This section emphasizes the careful balance between trusting AI-generated code and maintaining professional responsibility through layered security checks.</p>\n\n<p><strong>Reflections, Broader Implications, and Closing:</strong> In closing <a class=\"medius-aisummary\" data-starttime=\"00:10:05\">00:10:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>, Simon reflects on how AI has fundamentally shifted the landscape for professional developers, expanding what one person can achieve but also raising expectations. He demonstrates how he built Swift UI apps without prior experience, guided entirely by AI code assistants. He acknowledges using and trusting such auto-generated tools for months, citing accumulated user trust as more important than expert authorship. The hosts wrap up with praise for Simon’s pioneering work in AI-augmented software engineering, sharing a light moment around a query about Pelicans on his wildlife-blogging system <a class=\"medius-aisummary\" data-starttime=\"00:11:42\">00:11:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:55\">00:11:55</a>. The segment ends with gratitude to sponsors and viewers, closing out a session that highlights Simon Willison’s innovative integration of AI with decades of developer craftsmanship.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/75c37ec5-13dc-431b-89ea-c41f61c33148","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/75c37ec5-13dc-431b-89ea-c41f61c33148","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/75c37ec5-13dc-431b-89ea-c41f61c33148","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/75c37ec5-13dc-431b-89ea-c41f61c33148","speakerIds":["1768245294609build26-1769008130185001WAvh","1768245294609build26-1776365809006001iDqG"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/75c37ec5-13dc-431b-89ea-c41f61c33148","ogTitle":"Scott and Mark learn to Vibe Check with Simon Willison","ogDescription":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","deliveryTypes":[],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776365809006001iDqG","RegistrantKey":"1768245294609build26-1776365809006001iDqG"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"65038217-cf86-4260-a3cc-7171d79ac5a9","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Hanselman, Steve Sanderson","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780439116833001n1bf","langLocale":"en-US","sessionId":"1780439116833001n1bf","sessionInstanceId":"1780439116833001n1bf","sessionCode":"LIVE116","sessionCodeNormalized":"EN-US-LIVE116","title":"Scott and Mark learn to Vibe Check with Steve Sanderson","sortTitle":"scott and mark learn to vibe check with steve sanderson","description":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","aiDescription":"<p><strong>Introduction and Guest Reveal:</strong> The video begins with Scott and Mark welcoming viewers back and thanking the event sponsors <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>. They introduce themselves and announce their first guest, Steve Sanderson, a well-known software engineer recognized for his work on Knockout JS and Blazor <a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:18\">00:00:18</a>. The hosts build suspense by suggesting Steve will unveil something entirely new—perhaps created in a playful, AI-driven coding spree. This sets an enthusiastic tone for his presentation.</p>\n\n<p><strong>Setting the Stage: Evolution to Vibe OS:</strong> Steve opens by placing his demo in a historical context, playfully tracing human progression from stone tools and mathematics to modern computing <a class=\"medius-aisummary\" data-starttime=\"00:00:46\">00:00:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>. He humorously claims humanity has not yet reached its technological pinnacle—until his forthcoming reveal. He then introduces “Vibe OS,” the world’s first “hallucinated operating system,” suggesting it represents the next step beyond AI code generation <a class=\"medius-aisummary\" data-starttime=\"00:01:32\">00:01:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>. The audience is poised for something both absurd and groundbreaking.</p>\n\n<p><strong>Demo: Booting and Exploring the Hallucinated System:</strong> Steve proceeds to demonstrate Vibe OS, showing how it boots from a virtual hard drive and displays a sleek interface in Hyper-V <a class=\"medius-aisummary\" data-starttime=\"00:01:51\">00:01:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:09\">00:02:09</a>. He launches standard applications—Notepad, Calculator, and Internet Explorer—but reveals a key twist: none of these apps contain any code. Every interface element and behavior is hallucinated in real time by AI <a class=\"medius-aisummary\" data-starttime=\"00:02:44\">00:02:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:57\">00:02:57</a>. The apps work as expected, yet their logic and UI are completely fabricated, presenting a surreal view of computing where “everything is AI.”</p>\n\n<p><strong>Applications and Interactive Experiments:</strong> Continuing the demo, Steve runs searches and generates imagined applications within Vibe OS—everything from Encarta 98 reimagined with data about Mark Russinovich to fictitious software like a rude version of Commander XC <a class=\"medius-aisummary\" data-starttime=\"00:04:34\">00:04:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>. The playful hallucinations extend to Microsoft Money 95, paint apps that depict Scott Hanselman, and even nested operating systems simulating vintage computers and mobile environments <a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:00\">00:08:00</a>. The humor escalates as hosts react in disbelief, acknowledging the creativity and absurdity of the concept.</p>\n\n<p><strong>Technical Breakdown: How It Works:</strong> When asked to define “hallucination,” Steve explains the architecture behind Vibe OS <a class=\"medius-aisummary\" data-starttime=\"00:09:35\">00:09:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:22\">00:10:22</a>. Initially, he experimented with image generation but found it inefficient. He then moved to creating structured HTML UIs, leveraging Copilot SDK to simulate interface logic. Each window exists as an isolated iframe connecting to a session that outputs HTML, tracks clicks by element ID, and produces diffs to update the UI dynamically <a class=\"medius-aisummary\" data-starttime=\"00:10:01\">00:10:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>. The result is a deceptively interactive, fully AI-rendered interface—“intentional slop,” as the hosts dub it, demonstrating AI-assisted creativity taken to extremes.</p>\n\n<p><strong>Conclusion and Recognition:</strong> As the presentation wraps up, Scott and Mark praise Steve’s ingenuity and humor, calling his creation a blend of “vibes” and advanced AI software engineering <a class=\"medius-aisummary\" data-starttime=\"00:11:13\">00:11:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:20\">00:11:20</a>. They joke about how quickly it was built—within several hours—and present Steve with an award for his innovative and entertaining demonstration <a class=\"medius-aisummary\" data-starttime=\"00:11:23\">00:11:23</a>. The hosts then close by thanking him for the “amazing vibes” and transitioning to their sponsor segment, leaving the audience with a mix of laughter and inspiration about the boundary between reality and AI imagination.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/412eb4f9-6245-4b44-9497-dd6019195fe5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/412eb4f9-6245-4b44-9497-dd6019195fe5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/412eb4f9-6245-4b44-9497-dd6019195fe5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/412eb4f9-6245-4b44-9497-dd6019195fe5","speakerIds":["1768245294609build26-1769008130185001WAvh","1768245294609build26-1776185877803001f40v"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/412eb4f9-6245-4b44-9497-dd6019195fe5","ogTitle":"Scott and Mark learn to Vibe Check with Steve Sanderson","ogDescription":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","deliveryTypes":[],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776185877803001f40v","RegistrantKey":"1768245294609build26-1776185877803001f40v"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"7f9ba646-eb8a-458e-92fb-6479177831e9","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"swyx (Shawn Wang), Scott Hanselman","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780439084756001sJdk","langLocale":"en-US","sessionId":"1780439084756001sJdk","sessionInstanceId":"1780439084756001sJdk","sessionCode":"LIVE115","sessionCodeNormalized":"EN-US-LIVE115","title":"Scott and Mark learn to Vibe Check with Swyx","sortTitle":"scott and mark learn to vibe check with swyx","description":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","aiDescription":"<p><strong>Introduction and Context:</strong> At the start of the video <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>, Scott and Mark introduce a segment of \"Scott and Mark Learn to Vibe\" taking place live at Microsoft Build. They recap earlier demos including Steve Sanderson’s \"vibe coded OS\" and Cassidy Williams’ \"CSS SQLite database,\" before welcoming the next participant, Swix (Sean), known for his AI Engineer World’s Fair and his expertise in AI engineering. The hosts set the stage for Swix’s demonstration, expressing anticipation about what he might show and seeing if he can top the earlier presenters <a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>.</p>\n\n<p><strong>App Introduction and Purpose:</strong> Swix begins his presentation <a class=\"medius-aisummary\" data-starttime=\"00:00:31\">00:00:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:11\">00:01:11</a>, greeting the hosts and audience. He explains that instead of a playful demo, he’s showcasing a real internal app he built for work—a scheduling system for his engineering conference. He describes the app as something he’s slightly nervous to reveal since it’s under development but actively being used by his team. He runs one of the largest AI engineering conferences in San Francisco and shows an example page from last year, which had over 3,000 participants and numerous speakers and sponsors <a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>. He outlines how the event logistics involve managing 10 parallel tracks and thousands of sessions, and explains dissatisfaction with the external vendor tools used previously that lacked customization and visual appeal <a class=\"medius-aisummary\" data-starttime=\"00:01:44\">00:01:44</a>.</p>\n\n<p><strong>Design of the Scheduling System and AIE Bot:</strong> Swix moves into the new year’s conference plans <a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:47\">00:02:47</a>, which he says has doubled in size, with 6,000 attendees and even more sponsors and talks. To handle this scale, he built his own scheduling app called \"AIE Bot,\" designed to solve a “bin-packing problem,” where sessions and speakers must fit precisely into available days and tracks. He stresses precision due to the risk of placing speakers or attendees in incorrect slots. He demonstrates the app by scheduling Scott Hanselman into a specific track <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>, emphasizing how it proposes minimal changes while requiring human approval to maintain accuracy. He discusses the mix between agents and human oversight, combining automation with exact control to prevent AI hallucinations and errors.</p>\n\n<p><strong>Integration Across Platforms and Workflow:</strong> Expanding on functionality <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:07\">00:06:07</a>, Swix shows how AIE Bot connects to communication channels like Slack and email, allowing users to issue commands directly within those interfaces. He adds Mark Russinovich to an event track using Slack and describes plans for omnipresent agents that respond via text, Slack, and email. He demonstrates sending messages and receiving confirmation replies automatically processed by mini models, not large “frontier” LLMs, allowing for faster, cheaper interactions. The discussion transitions to technical aspects: his agents run on Cloudflare using a custom-coded while loop instead of an SDK, giving him tight control over bounded autonomy in agents. He emphasizes how avoiding unlimited autonomy improves safety and predictability, and admits the entire system was \"vibe coded\" by AI tools.</p>\n\n<p><strong>Technical Details and Architecture:</strong> The conversation shifts deeper into implementation <a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>. Swix explains he used Cloudflare’s Wrangler CLI for deployment and GitHub Actions for database migrations and continuous deployment. The backend uses Cloudflare’s D1 database with an audit and rollback system for precision edits. He discusses the advantage of fully custom software versus the SaaS products costing up to $200,000 per year, noting his bespoke version fits his organization’s needs perfectly. While specialized for his event dates and structure, he acknowledges the tool isn’t general-purpose—and though it could be generalized, he prefers focusing on productivity rather than commercialization <a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>.</p>\n\n<p><strong>Future Possibilities and Conclusion:</strong> As the demo wraps up <a class=\"medius-aisummary\" data-starttime=\"00:09:35\">00:09:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>, the group discusses potential next features like adding voice integration. Swix argues that voice should be handled at the operating system level—not baked individually into each app—to achieve global optimization. They exchange thoughts on using system-level keys (like Windows key H) for consistent access to voice agents. Finally, Swix confirms the project is deployed through GitHub Actions and entirely “vibe coded”—built almost entirely by AI with his guidance and feedback loops within Cursor IDE. He mentions maintaining parallelism and controlled autonomy among agents to ensure maintainability. The hosts praise the app’s functionality and creativity before closing the segment with applause and transitioning to sponsor messages.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e52b8eb8-f1d1-4659-9e59-f1754039b9a1","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e52b8eb8-f1d1-4659-9e59-f1754039b9a1","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e52b8eb8-f1d1-4659-9e59-f1754039b9a1","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e52b8eb8-f1d1-4659-9e59-f1754039b9a1","speakerIds":["1768245294609build26-1775516734838001pZmI","1768245294609build26-1769008130185001WAvh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e52b8eb8-f1d1-4659-9e59-f1754039b9a1","ogTitle":"Scott and Mark learn to Vibe Check with Swyx","ogDescription":"AI can turn an idea into a working demo faster than ever. But can that demo survive two experts who have seen every trick in the book? In this live Build showcase, developers present AI-assisted apps, agents, tools, and workflows to Mark Russinovich and Scott Hanselman. Mark and Scott will ask how it works, where the seams are, what the AI actually built, and whether the result is clever prototype, production-ready software, or something unexpectedly magical. Come for the demos. Stay for the technical reveal.","deliveryTypes":[],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775516734838001pZmI","RegistrantKey":"1768245294609build26-1775516734838001pZmI"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"48c793f5-2987-44e4-9813-5eb7d675f101","startDateTime":"2026-06-03T00:25:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Fernanda Herger, Priyanka Shah, Betsy Weber, Jeremy Sinclair","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780001402064001ifTy","langLocale":"en-US","sessionId":"1780001402064001ifTy","sessionInstanceId":"1780001402064001ifTy","sessionCode":"LIVE176","sessionCodeNormalized":"EN-US-LIVE176","title":"So you want to become an MVP?","sortTitle":"so you want to become an mvp?","description":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.","aiDescription":"<p><strong>Introduction and Team Overview:</strong> The video begins with a greeting at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, introducing representatives from the Microsoft MVP (Most Valuable Professional) program. Fernando Sariva, a Program Manager focused on community engagement, introduces himself and his colleague Betsy Weber, Senior Engagement Lead with nearly ten years on the MVP team. They are joined by MVPs Jeremy Sinclair and Priyanka, who represent different technology domains and geographic locations, showcasing the program’s global scope (<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>).</p>\n\n<p><strong>MVP Program Overview and Global Reach:</strong> Fernando and Betsy explain the foundation of the MVP program—a recognition initiative that has existed for over 30 years, honoring outstanding technology community leaders. As of recording, the program includes over 4000 MVPs across 96 countries speaking 49 languages (<a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:24\">00:01:24</a>). Betsy elaborates that being an MVP means freely sharing knowledge and helping Microsoft improve its products. The hosts then transition to hearing about personal experiences from Jeremy and Priyanka, illustrating how MVP recognition has influenced their professional growth (<a class=\"medius-aisummary\" data-starttime=\"00:01:24\">00:01:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>).</p>\n\n<p><strong>Professional Impact and Community Building:</strong> Jeremy explains that being part of the MVP community has opened opportunities with various companies that noticed his open-source contributions (<a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:21\">00:02:21</a>). While he focuses on interactions through social platforms such as LinkedIn, X, GitHub, and BlueSky, Priyanka shares that running user groups helps create local forums for technology learning and collaboration, especially in fast-evolving fields like AI (<a class=\"medius-aisummary\" data-starttime=\"00:03:12\">00:03:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:27\">00:03:27</a>). She describes organizing events like “Agent Con” and “Agent Camp” under the Global AI Community banner (<a class=\"medius-aisummary\" data-starttime=\"00:03:52\">00:03:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:56\">00:05:56</a>), where participants learn about and build agentic frameworks—helping bridge theory and enterprise application. The conversation emphasizes how these efforts evolve technology understanding and encourage collaboration across industry ecosystems.</p>\n\n<p><strong>Balancing Passion, Contribution, and Recognition:</strong> The discussion shifts as Fernando asks how MVPs balance their work and personal lives while maintaining high engagement levels (<a class=\"medius-aisummary\" data-starttime=\"00:06:25\">00:06:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>). Jeremy explains that coding and open-source contributions are both his decompression and passion. He recounts porting PowerToys to ARM64 for his Surface Pro X (<a class=\"medius-aisummary\" data-starttime=\"00:07:44\">00:07:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:04\">00:08:04</a>), an effort that expanded collaboration with Microsoft engineers and deepened his skills. Betsy then highlights key benefits such as direct interaction with Microsoft product teams, early previews, and insider insights (<a class=\"medius-aisummary\" data-starttime=\"00:08:53\">00:08:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:05\">00:09:05</a>). Jeremy shares that the “badge of honor” comes with personal doubts and growth—each renewal motivates him to contribute more and embrace his role in shaping technology feedback.</p>\n\n<p><strong>Inspiration to Engage and Career Transformation:</strong> Priyanka explains what sparked her journey into community work—starting with informal coffee chats about machine learning, then transitioning into structured meetups and evangelism programs (<a class=\"medius-aisummary\" data-starttime=\"00:11:26\">00:11:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:59\">00:13:59</a>). She encourages aspiring MVPs to express their enthusiasm through any medium—public speaking, coding, blogging, or organizing local events—and emphasizes that passion, not perfection, drives recognition. Betsy reinforces that individuals should follow their unique way of giving back rather than trying to “do it all” (<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:13\">00:14:13</a>). Later, Jeremy shares that the MVP honor enhanced his professional confidence and public presence—he proudly represents West Virginia’s only MVP (<a class=\"medius-aisummary\" data-starttime=\"00:14:28\">00:14:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:19\">00:15:19</a>). Priyanka adds that the MVP title builds client trust and strengthens her mentoring role in helping others adapt to the fast-changing AI landscape (<a class=\"medius-aisummary\" data-starttime=\"00:15:38\">00:15:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:39\">00:16:39</a>).</p>\n\n<p><strong>Closing Reflections and Advice to Future MVPs:</strong> As the video concludes (<a class=\"medius-aisummary\" data-starttime=\"00:17:36\">00:17:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:40\">00:18:40</a>), Jeremy defines being an MVP as a continuous journey of personal and professional growth, fostering connections while giving back to the tech community. Priyanka emphasizes that the MVP award itself isn’t the end goal—the real reward lies in sustained passion and contribution. Fernando and Betsy thank both guests, inviting viewers to explore more at the official MVP learning page (<a class=\"medius-aisummary\" data-starttime=\"00:18:45\">00:18:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:53\">00:18:53</a>). The discussion closes with a united message: stay passionate, contribute authentically, and keep learning through community engagement.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"00:25 - 00:40","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:40:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a0766ee6-01ec-4b37-a1b3-ce1ec4426411","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a0766ee6-01ec-4b37-a1b3-ce1ec4426411","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a0766ee6-01ec-4b37-a1b3-ce1ec4426411","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a0766ee6-01ec-4b37-a1b3-ce1ec4426411","speakerIds":["1768245294609build26-1776209926960001eLK3","1768245294609build26-1773185276975001cATe","1768245294609build26-1758299331835001JwKs","1768245294609build26-1772566305259001tYa6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a0766ee6-01ec-4b37-a1b3-ce1ec4426411","ogTitle":"So you want to become an MVP?","ogDescription":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776209926960001eLK3","RegistrantKey":"1768245294609build26-1776209926960001eLK3"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1772566305259001tYa6","RegistrantKey":"1768245294609build26-1772566305259001tYa6"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758299331835001JwKs","RegistrantKey":"1768245294609build26-1758299331835001JwKs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773185276975001cATe","RegistrantKey":"1768245294609build26-1773185276975001cATe"}],"repeatedSessions":[],"relatedResources":[{"link":"https://mvp.microsoft.com/","backpackLink":null,"label":"So you want to become an MVP?","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"9370f808-e519-4197-8cd2-5f9e1b529384","startDateTime":"2026-06-03T00:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alycea Adams, Tom Davis","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780002083743001O0kt","langLocale":"en-US","sessionId":"1780002083743001O0kt","sessionInstanceId":"1780002083743001O0kt","sessionCode":"LIVE142","sessionCodeNormalized":"EN-US-LIVE142","title":"Student to Startup","sortTitle":"student to startup","description":"Alycea Adams, founder and Imagine Cup World Championship finalist, joins Tom Davis, GM of Microsoft for Startups, to talk about how Microsoft supports founders at the moments that matter, starting at the student stage. They explore why Imagine Cup is such a strong entry point for early-stage founders ready to move beyond the idea phase, connecting them to the tools, guidance, and ecosystem needed to build and grow.","aiDescription":"<p><strong>Introduction and Background:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, the interviewer introduces Alicia, the CEO and founder of Hair Match, an Imagine Cup finalist. Alicia explains that Hair Match is a mobile app allowing users, primarily those with textured hair, to upload photos and receive personalized product recommendations. She notes that the idea came from her own experiences struggling to find suitable hair products and wanting to streamline that process through technology.</p>\n\n<p><strong>Student Founder Journey:</strong> Between <a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>, Alicia shares how she launched Hair Match as a student, balancing academics, athletics, and entrepreneurship. A cheerleader and information science major with minors in entrepreneurship and urban planning, she emphasizes having built the skills to handle multiple priorities. She describes turning her personal experience and her online community’s engagement into motivation to test and validate the Hair Match concept, taking advantage of the flexibility and risk tolerance she had as a young founder.</p>\n\n<p><strong>Technology and Microsoft Partnership:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a> through <a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>, Alicia discusses her academic and technical background and how she connected with Microsoft. She recounts meeting her co-founder during an internship at Delta Airlines and learning how to use Microsoft technologies despite not being a developer. Through the Imagine Cup and Microsoft for Startups, Hair Match received mentorship and Azure credits to build their prototype, leveraging design thinking and AI services. The program gave her tools and support to turn her idea into a sustainable startup.</p>\n\n<p><strong>Growth, AI Integration, and Product Development:</strong> Between <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>, Alicia speaks about Hair Match’s evolution over the past year. She reflects on embracing moments of calm to focus on iterative improvement and user feedback. She explains using Azure OpenAI and AI Foundry models to power the app’s photo recognition capability—allowing users to analyze their hair and receive accurate suggestions. Alicia sees immense potential for vision-language models beyond hair care, emphasizing how accessible AI tools empower non-technical founders to innovate. She also touches on her pivot toward research and development partnerships with major beauty retailers to improve data-driven product creation.</p>\n\n<p><strong>Business Model and Support from Microsoft:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:07:06\">00:07:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:58\">00:08:58</a>, Alicia outlines Hair Match’s subscription-based model. For now, the app is free to build its user base and gather data that will help improve product recommendations and formulations. She credits Microsoft for Startups and the Imagine Cup for providing not just technology but mentorship in pitching, investor relations, and business strategy. Alicia emphasizes how Microsoft’s programs help founders overcome common hurdles and foster a collaborative “family” among young innovators.</p>\n\n<p><strong>User Insights and Future Vision:</strong> In the final section, <a class=\"medius-aisummary\" data-starttime=\"00:09:09\">00:09:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:13\">00:13:13</a>, Alicia reveals that her expected audience of young individual users evolved into mothers seeking solutions for their daughters’ hair. She attributes this discovery to using AI models for fine-tuned data collection and multi-shot prompting. Discussing future plans, she intends to grow Hair Match into a data repository for textured hair research to assist product manufacturers. Reflecting personally, Alicia encourages other founders to give themselves grace to make mistakes, embrace learning, and speak their ideas aloud to build momentum. The conversation ends with the interviewer praising her achievements and wishing her continued success, as Alicia expresses gratitude for the support and inspiration she’s received through Microsoft’s programs.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"00:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b2a624c7-5cb0-4a3d-b252-274a14cef87e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b2a624c7-5cb0-4a3d-b252-274a14cef87e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b2a624c7-5cb0-4a3d-b252-274a14cef87e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b2a624c7-5cb0-4a3d-b252-274a14cef87e","speakerIds":["1768245294609build26-1778004187096001mqFQ","1768245294609build26-1762363782373001CUXO"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b2a624c7-5cb0-4a3d-b252-274a14cef87e","ogTitle":"Student to Startup","ogDescription":"Alycea Adams, founder and Imagine Cup World Championship finalist, joins Tom Davis, GM of Microsoft for Startups, to talk about how Microsoft supports founders at the moments that matter, starting at the student stage. They explore why Imagine Cup is such a strong entry point for early-stage founders ready to move beyond the idea phase, connecting them to the tools, guidance, and ecosystem needed to build and grow.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778004187096001mqFQ","RegistrantKey":"1768245294609build26-1778004187096001mqFQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762363782373001CUXO","RegistrantKey":"1768245294609build26-1762363782373001CUXO"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"5b242c03-5d9c-430d-8b47-889f83b0e036","startDateTime":"2026-06-03T21:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Harald Kirschner","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780009803750001jqup","langLocale":"en-US","sessionId":"1780009803750001jqup","sessionInstanceId":"1780009803750001jqup","sessionCode":"LIVE169","sessionCodeNormalized":"EN-US-LIVE169","title":"The New Agents Window in VS Code","sortTitle":"the new agents window in vs code","description":"Agent sessions are more powerful than ever and the VS Code team has been building the visibility and efficiency to match. Harald Kirschner walks through the new Agents window, the token optimization work shipping across recent releases, and how to get the most out of every session.","aiDescription":"<p><strong>Introduction and Team Overview:</strong> The discussion opens at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a> with Harold joining from the Visual Studio Code (VS Code) team. The host notes that nearly the entire VS Code team is present, mentioning members like Courtney, Joanna, Josh, and Tyler. Harold confirms his role as a \"PM-ish\" on the team, explaining that VS Code's project management culture is highly technical, with all new members—whether PMs, advocates, or documentation writers—required to create code extensions when they join <a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>. This sets the stage for a deep technical conversation about recent developments within VS Code.</p>\n\n<p><strong>Introducing the Agent Window Concept:</strong> The talk transitions at <a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a> to focus on the new \"agent window,\" previously called the \"agent sessions window.\" Harold describes how it evolved to streamline work with multiple coding agents and projects. Traditionally, developers switched between numerous VS Code windows, causing mental context loss and workflow overhead. The agent window unifies all open workspaces and agent processes—local or remote—into a single, centralized interface, improving continuous development and orchestration <a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>. Although it looks and feels like standard VS Code, the agent window integrates all extensions and themes, operating as a redesigned form factor within the same VS Code application environment.</p>\n\n<p><strong>Demonstration of Multi-Workspace Capabilities:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:03:05\">00:03:05</a>, Harold demonstrates setting up agents across different repositories, explaining that workspaces can be local folders or remote ones hosted via GitHub or dev tunnels. He elaborates that VS Code tunnels allow users to open a web-accessible URL to their development environment on any machine <a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>, emphasizing it's a powerful but often overlooked feature of VS Code. This integration makes the agent window effective for developers juggling multiple open-source projects or pending pull requests, enabling real-time tracking of progress and context across environments. Harold describes the interface as an \"agent orchestration control center\" where all sessions—from docs to source code repos—appear in collapsible panes, promoting better task organization.</p>\n\n<p><strong>Customization and Agent Feedback:</strong> By <a class=\"medius-aisummary\" data-starttime=\"00:05:28\">00:05:28</a>, the conversation moves to how developers can manage notifications and customizations when agents complete tasks. Harold explains hooks for auditory feedback when an agent finishes running, addressing challenges of excessive alerts like constant \"ding\" sounds from concurrent agents. He shares his setup using VS Code customization hooks to trigger spoken summaries through local speech models when a task completes—his computer announces completion details aloud <a class=\"medius-aisummary\" data-starttime=\"00:06:33\">00:06:33</a>. This clever integration showcases how agent notifications can be made more intuitive and context-aware, bridging functionality from command-line autopilot operations with personalized system feedback.</p>\n\n<p><strong>Cost Optimization and Model Routing:</strong> Near the session’s close at <a class=\"medius-aisummary\" data-starttime=\"00:07:31\">00:07:31</a>, Harold introduces cost and performance improvements tied to new model usage in VS Code’s AI features. He explains that the “auto” model now routes requests dynamically based on task complexity, delivering roughly a 10% discount in token usage compared to fixed model runs. The recommended approach is to plan tasks using larger models before letting the auto model handle implementation, which smartly selects the most efficient computational route <a class=\"medius-aisummary\" data-starttime=\"00:08:13\">00:08:13</a>. This optimization helps developers saving tokens—essentially computing credits—while maintaining output quality. Future updates aim to display the exact token savings retrospectively, similar to retail-style \"You saved X today\" feedback, enhancing the sense of user efficiency and success.</p>\n\n<p><strong>Closing Remarks:</strong> The conversation wraps up around <a class=\"medius-aisummary\" data-starttime=\"00:09:24\">00:09:24</a> with Harold reminding viewers to check release notes for ongoing improvements in pricing models and token savings. The host thanks Harold for his detailed explanations and insights, signaling the end of the session <a class=\"medius-aisummary\" data-starttime=\"00:09:29\">00:09:29</a>. The video concludes on an appreciative note, acknowledging the blend of engineering depth and user-focused innovation within the VS Code ecosystem.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:00 - 21:10","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:10:00Z","durationInMinutes":10,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/39e7b927-a96f-4e0c-bc26-8d9fb757cd58","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/39e7b927-a96f-4e0c-bc26-8d9fb757cd58","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/39e7b927-a96f-4e0c-bc26-8d9fb757cd58","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/39e7b927-a96f-4e0c-bc26-8d9fb757cd58","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1774285000231001pWGh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/39e7b927-a96f-4e0c-bc26-8d9fb757cd58","ogTitle":"The New Agents Window in VS Code","ogDescription":"Agent sessions are more powerful than ever and the VS Code team has been building the visibility and efficiency to match. Harald Kirschner walks through the new Agents window, the token optimization work shipping across recent releases, and how to get the most out of every session.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000231001pWGh","RegistrantKey":"1768245294609build26-1774285000231001pWGh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/Agentswindow","backpackLink":null,"label":"Meet the Agents window: a dedicated space for orchestrating agent sessions across all your projects.","category":"Web link"},{"link":"https://vscodepodcast.com/22","backpackLink":null,"label":"Go deeper on the new Agents window with the recent VS Code Insiders Podcast episode","category":"Web link"},{"link":"https://aka.ms/VSCode/TTCompression","backpackLink":null,"label":"Use fewer tokens in your agent sessions with VS Code.","category":"Web link"},{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Build your VS Code and Copilot skills with free, hands-on courses.","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"ce322677-c7f9-4df5-b665-c8ea2959c1e6","startDateTime":"2026-06-03T19:55:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Burke Holland, Courtney Webster","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780008693994001e0F4","langLocale":"en-US","sessionId":"1780008693994001e0F4","sessionInstanceId":"1780008693994001e0F4","sessionCode":"LIVE167","sessionCodeNormalized":"EN-US-LIVE167","title":"The PM/dev handoff isn't what it used to be","sortTitle":"the pm/dev handoff isn't what it used to be","description":"The PM to dev handoff used to mean a document nobody loved reading and a long wait to find out if the idea actually worked. When agents can take a vague idea to a working pull request in a day, the whole loop changes. This session is about what that means for how product and engineering work together now and what a faster, prototype-first culture actually looks like on a team shipping weekly.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"19:55 - 20:10","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:10:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/50305603-5a3c-4f95-ba05-5169880e9354","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/50305603-5a3c-4f95-ba05-5169880e9354","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/50305603-5a3c-4f95-ba05-5169880e9354","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/50305603-5a3c-4f95-ba05-5169880e9354","speakerIds":["1768245294609build26-1776347378057001Rvny","1768245294609build26-1774284995211001pJD3"],"sessionLinks":[],"relatedSessionCodes":["BRK204"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/50305603-5a3c-4f95-ba05-5169880e9354","ogTitle":"The PM/dev handoff isn't what it used to be","ogDescription":"The PM to dev handoff used to mean a document nobody loved reading and a long wait to find out if the idea actually worked. When agents can take a vague idea to a working pull request in a day, the whole loop changes. This session is about what that means for how product and engineering work together now and what a faster, prototype-first culture actually looks like on a team shipping weekly.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995211001pJD3","RegistrantKey":"1768245294609build26-1774284995211001pJD3"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/AIBlog","backpackLink":null,"label":"Read the blog How the VS Code team actually builds with AI, including the PM PR workflow and what went into going weekly","category":"Web link"},{"link":"https://aka.ms/VSCode/Build26/BRK204","backpackLink":null,"label":"Watch (on demand) the Build breakout session on How the VS Code team builds with AI","category":"Web link"},{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Grow your VS Code and GitHub Copilot skills with free, hands-on courses","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"746edb08-d9ae-4bd4-9c53-f36dd7dd173e","startDateTime":"2026-06-03T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Marco Casalaina","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780010477957001fyyC","langLocale":"en-US","sessionId":"1780010477957001fyyC","sessionInstanceId":"1780010477957001fyyC","sessionCode":"LIVE171","sessionCodeNormalized":"EN-US-LIVE171","title":"The Three IQs: Ground Your Agents in Knowledge, Data, and Work","sortTitle":"the three iqs: ground your agents in knowledge, data, and work","description":"Agents shouldn't hold all the context themselves — they should delegate. See how Foundry IQ (knowledge), Fabric IQ (data), and Work IQ (human context) let developers build grounded, enterprise-ready agents faster, with reusable intelligence instead of hand-wired pipelines. Live demo included.","aiDescription":"<p><strong>Introduction and Context:</strong> The video opens with greetings and introductions from Marco Caslena, VP of Products at Core AI, and Aisha, a Developer Advocate at Microsoft (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>). They introduce the topic — the “Four IQs,” a concept expanding from previously three IQs to four, which represent different capabilities for intelligent agents. The discussion begins by setting the stage at Microsoft Build, establishing IQs as a major topic and contextualizing them as components needed for real-world agent operations with complex data sources (<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>).</p>\n\n<p><strong>Overview of the Four IQs:</strong> Aisha and Marco outline the four distinct IQs — Web IQ, Foundry IQ, Fabric IQ, and Work IQ (<a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:58\">00:00:58</a>). Web IQ connects agents to web-based knowledge for accessing online data and services such as shipment tracking or local information (<a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>). Work IQ integrates Microsoft 365 data like emails and Teams messages, enabling agents to interact with organizational contexts. Fabric IQ manages structured data, providing headless access for agents to structured sources such as Power BI reports. Foundry IQ handles unstructured data repositories — blob storage, SharePoint, and search indices — with advanced retrieval mechanisms beyond basic retrieval augmented generation (<a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>).</p>\n\n<p><strong>Early Development and Agent Instruction Optimization:</strong> Marco and Aisha describe the process and challenges of integrating these IQs into working agents. While the individual IQ tools are new, teams are collaborating to align them effectively (<a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:18\">00:03:18</a>). They emphasize the importance of defining precise agent instructions to ensure optimal decision-making and tool selection. The conversation dives into the iteration of instruction sets and mappings for each IQ, including examples of JSON body requests and search or action behaviors tailored to Work IQ. This detailed configuration allows agents to consume IQ content appropriately and achieve high performance, not just basic functionality (<a class=\"medius-aisummary\" data-starttime=\"00:03:30\">00:03:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:48\">00:04:48</a>).</p>\n\n<p><strong>Preventing Context Redundancy and Introducing Agent Templates:</strong> The next section discusses how IQs help reduce “context rot,” meaning developers no longer need to overload agents with excessive contextual data (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:17\">00:05:17</a>). Even with IQs, defining the specific instructions for when and how to use each tool remains a developer’s task. Marco demonstrates the concept of “agent templates” — predefined agent structures that allow anyone to instantiate their own version within Microsoft 365 (<a class=\"medius-aisummary\" data-starttime=\"00:06:03\">00:06:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:11\">00:06:11</a>). Each instantiated agent possesses unique identity attributes like its own inbox and Teams presence. Marco illustrates this using a “refund processor agent” example that responds to emails in its own security context, reinforcing the practical value of dedicated agent identities (<a class=\"medius-aisummary\" data-starttime=\"00:06:32\">00:06:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>).</p>\n\n<p><strong>Security Context and Organizational Deployment Considerations:</strong> Marco continues by explaining how agent identity contributes to secure organizational deployment (<a class=\"medius-aisummary\" data-starttime=\"00:07:31\">00:07:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:41\">00:08:41</a>). In contrast with “on behalf of” agents that act as users, certain agents should have separate permissions for cybersecurity reasons. He demonstrates this concept using an internal application, Copilot CLI, and shows how an agent that operates independently can have limited privileges (e.g., can send messages but cannot delete files). This design ensures that agent actions remain compliant and contained within their security posture, essential for enterprise-scale deployment.</p>\n\n<p><strong>Demo Recap, Resources, and Closing:</strong> Aisha concludes the video by showing trace logs that capture synchronized interaction among the IQs — Work IQ, Fabric IQ, and Foundry IQ — as they collaboratively processed a refund case for a customer (<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:22\">00:09:22</a>). She encourages viewers to explore the dedicated IQ learning series at “aka.ms/IQ-series,” where each IQ is covered in depth, with Foundry, Work, and Fabric IQ episodes available and Web IQ to follow (<a class=\"medius-aisummary\" data-starttime=\"00:09:41\">00:09:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:04\">00:10:04</a>). Completing all modules and cookbooks grants participants a digital badge acknowledging their expertise. Marco closes by urging everyone to “use yourself some IQs” and wishes the audience a good day (<a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:13\">00:10:13</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"21:30 - 21:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/eaaffcc3-d8a7-47ab-a42c-d84a91d007e7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/eaaffcc3-d8a7-47ab-a42c-d84a91d007e7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/eaaffcc3-d8a7-47ab-a42c-d84a91d007e7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/eaaffcc3-d8a7-47ab-a42c-d84a91d007e7","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1758304897855001h9Qq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/eaaffcc3-d8a7-47ab-a42c-d84a91d007e7","ogTitle":"The Three IQs: Ground Your Agents in Knowledge, Data, and Work","ogDescription":"Agents shouldn't hold all the context themselves — they should delegate. See how Foundry IQ (knowledge), Fabric IQ (data), and Work IQ (human context) let developers build grounded, enterprise-ready agents faster, with reusable intelligence instead of hand-wired pipelines. Live demo included.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758304897855001h9Qq","RegistrantKey":"1768245294609build26-1758304897855001h9Qq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/iq-series","backpackLink":null,"label":"The Three IQs: Ground Your Agents in Knowledge, Data, and Work","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"5088af56-931a-46d6-afcd-6b8174158210","startDateTime":"2026-06-03T19:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Atul Aggarwal, Seth Juarez, Linda Li","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780007527401001m8Ql","langLocale":"en-US","sessionId":"1780007527401001m8Ql","sessionInstanceId":"1780007527401001m8Ql","sessionCode":"LIVE163","sessionCodeNormalized":"EN-US-LIVE163","title":"Toolboxes in Foundry: Build, Search and Govern Your Tools","sortTitle":"toolboxes in foundry: build, search and govern your tools","description":"Today, each agent often wires tools directly, with its own authentication, credentials, and integration code. Toolboxes in Foundry are a new way to build, search and govern all tool types across all of your AI agents without rewiring them every time, powered by runtime tool search, unified endpoint and guardrail integration.","aiDescription":"<p><strong>Introduction and Setup:</strong> The video opens with Seth welcoming viewers back to the Build stage and introducing his colleagues Linda and Atul <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>. Seth admits that earlier, when Linda mentioned the new Microsoft Foundry Toolbox, he was skeptical about its utility. However, he confesses that he was proven wrong because the Toolbox played a key role in the keynote demo they delivered successfully <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>. Linda introduces herself as a product manager at Microsoft Foundry, and Atul describes his role as a partner engineering manager <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:50\">00:00:50</a>. Seth then invites them to discuss the main problems that Toolbox aims to solve.</p>\n\n<p><strong>Challenges Addressed by Toolbox:</strong> Linda outlines the key issues developers face when building AI agents at scale: authentication complexity, multiplicity of tools, and governance <a class=\"medius-aisummary\" data-starttime=\"00:00:58\">00:00:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:21\">00:02:21</a>. As developers connect agents to many tools (often hundreds), each with different APIs and protocols, they must write extensive custom code for authentication and authorization. Additionally, exposing too many tools to the model inflates token usage and clutters context windows. Governance adds more difficulty, since each tool may require unique policy enforcement, making code reuse and scaling inefficient. Atul further explains that Toolbox tackles these problems by providing a unified structure for tool management, policy application, and context optimization <a class=\"medius-aisummary\" data-starttime=\"00:02:22\">00:02:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:43\">00:02:43</a>.</p>\n\n<p><strong>How Toolbox Works for Developers:</strong> Atul demonstrates how easy it is to start using Toolbox, showing that developers first curate their suite of required tools and then plug them into the Toolbox <a class=\"medius-aisummary\" data-starttime=\"00:02:45\">00:02:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>. Toolbox outputs a unified MCP-compatible endpoint that agents can call directly, simplifying tool invocation. This design reduces context bloat and token cost because Toolbox smartly routes each query to the correct tool without exposing unnecessary metadata to the model. Governance guardrails from Foundry can also be linked to Toolbox for automatic enforcement during every invocation, ensuring data safety. Seth highlights how this setup prevents procedural memory pollution within the LLM context space and asks about search capabilities within Toolbox.</p>\n\n<p><strong>Tool Search and Automation:</strong> Linda elaborates on Toolbox’s \"tool search\" capability, which dynamically retrieves the most relevant tools based on a user’s prompt <a class=\"medius-aisummary\" data-starttime=\"00:04:51\">00:04:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:14\">00:06:14</a>. Users can also \"ping\" tools to ensure they remain accessible to the agent at all times or let Toolbox automatically learn usage patterns and \"auto-ping\" frequently used tools. Seth and Atul clarify that when tool search is active, only two meta tools—search and call—are exposed to the model, greatly simplifying memory footprint. This design further supports efficient procedural control while still enabling parameter bindings and preconfigured credentials for tools, which can be set up just once by developers <a class=\"medius-aisummary\" data-starttime=\"00:06:35\">00:06:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>. Developers can also access individual tools or create combinations within Toolbox as needed.</p>\n\n<p><strong>Demo and Advanced Features:</strong> Linda presents a live Toolbox demo that includes a custom MCP server for inventory lookup and other tools such as knowledge bases and APIs <a class=\"medius-aisummary\" data-starttime=\"00:08:40\">00:08:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:24\">00:10:24</a>. Toolbox supports multiple custom combinations, letting users build multiple boxes for different tasks. Linda explains that skills complement tools by teaching agents \"how\" to use them, addressing customer needs like private skill catalogs and auto-discovery features. Seth realizes that Toolbox’s integration of both tools and skills provides stronger procedural guidance for LLMs. Linda shows how to attach a Toolbox as an MCP endpoint to any agent framework, reducing thousands of lines of integration code <a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:04\">00:12:04</a>. Atul adds that this approach enables multiplexing, where multiple agents can share optimized tool collections with minimal processing overhead.</p>\n\n<p><strong>Optimization, Future Plans, and Closing:</strong> The conversation concludes with a deeper dive into Toolbox’s internal mechanisms. Atul explains that Toolbox uses algorithms like BM25 for tool search today and will introduce LLM-enhanced search soon <a class=\"medius-aisummary\" data-starttime=\"00:12:26\">00:12:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:53\">00:12:53</a>. Seth compares the system to retrieval-augmented generation (RAG), noting that Toolbox operates like RAG but for procedural tools instead of documents. Linda announces that Toolbox is currently in public preview and will achieve general availability later this month <a class=\"medius-aisummary\" data-starttime=\"00:13:17\">00:13:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:34\">00:13:34</a>. New governance dashboards will soon allow customers to monitor tool usage, token consumption, ROI, and security controls, minimizing risks like prompt injection and data leakage. Seth wraps up by applauding Linda and Atul for their work on this innovative solution and thanks viewers for joining the session <a class=\"medius-aisummary\" data-starttime=\"00:13:57\">00:13:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"19:15 - 19:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:30:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ad613b19-51cf-4c92-a4c5-cae4670df3ae","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ad613b19-51cf-4c92-a4c5-cae4670df3ae","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ad613b19-51cf-4c92-a4c5-cae4670df3ae","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ad613b19-51cf-4c92-a4c5-cae4670df3ae","speakerIds":["1768245294609build26-1776709857973001bLgd","1768245294609build26-1751379013075001HHpX","1768245294609build26-1765790670451001e9yH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ad613b19-51cf-4c92-a4c5-cae4670df3ae","ogTitle":"Toolboxes in Foundry: Build, Search and Govern Your Tools","ogDescription":"Today, each agent often wires tools directly, with its own authentication, credentials, and integration code. Toolboxes in Foundry are a new way to build, search and govern all tool types across all of your AI agents without rewiring them every time, powered by runtime tool search, unified endpoint and guardrail integration.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776709857973001bLgd","RegistrantKey":"1768245294609build26-1776709857973001bLgd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765790670451001e9yH","RegistrantKey":"1768245294609build26-1765790670451001e9yH"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/Foundry/Toolbox/Docs","backpackLink":null,"label":"Try it now Toolbox is in public preview. Start here","category":"Web link"},{"link":"https://aka.ms/toolbox-build-blog","backpackLink":null,"label":"Read the announcement The full Toolbox intro post with the four pillars, setup walkthrough, and what's coming next","category":"Web link"},{"link":"https://aka.ms/Foundry/AgentService","backpackLink":null,"label":"Explore Foundry Agent Service The broader platform Toolbox lives in, including hosted agents, memory, and observability","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"017337a0-686a-423d-afac-4b7513034e11","startDateTime":"2026-06-03T18:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Seth Juarez, Daniel Rosenwasser","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780007260969001kKgn","langLocale":"en-US","sessionId":"1780007260969001kKgn","sessionInstanceId":"1780007260969001kKgn","sessionCode":"LIVE155","sessionCodeNormalized":"EN-US-LIVE155","title":"TypeScript updates: Latest Features and What's next","sortTitle":"typescript updates: latest features and what's next","description":"From improved type inference to performance wins, TypeScript keeps evolving. Daniel Rosenwasser joins Seth Juarez to walk through what's new in the language and what the team is building next.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"18:15 - 18:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:30:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e1161aa1-c289-4c29-8355-7f18e65827fe","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e1161aa1-c289-4c29-8355-7f18e65827fe","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e1161aa1-c289-4c29-8355-7f18e65827fe","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e1161aa1-c289-4c29-8355-7f18e65827fe","speakerIds":["1768245294609build26-1751379013075001HHpX","1768245294609build26-1776901346257001NT6h"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e1161aa1-c289-4c29-8355-7f18e65827fe","ogTitle":"TypeScript updates: Latest Features and What's next","ogDescription":"From improved type inference to performance wins, TypeScript keeps evolving. Daniel Rosenwasser joins Seth Juarez to walk through what's new in the language and what the team is building next.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776901346257001NT6h","RegistrantKey":"1768245294609build26-1776901346257001NT6h"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/TypeScript/NativePreview","backpackLink":null,"label":"Try TypeScript Native Preview","category":"Web link"},{"link":"https://aka.ms/TypeScript-6-0","backpackLink":null,"label":"TypeScript 6.0 announcement","category":"Web link"},{"link":"https://aka.ms/TypeScript-7-0/beta","backpackLink":null,"label":"Read the TypeScript 7.0 Beta announcement","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"05dac5c1-715a-45d6-99cb-063063b19721","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Evan Boyle, Burke Holland","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780003914870001QMOI","langLocale":"en-US","sessionId":"1780003914870001QMOI","sessionInstanceId":"1780003914870001QMOI","sessionCode":"LIVE152","sessionCodeNormalized":"EN-US-LIVE152","title":"What's new in GitHub Copilot CLI?","sortTitle":"what's new in github copilot cli?","description":"This session walks through what's new, including a redesigned terminal interface, Rubber Duck for second opinions, recurring prompts with /every, and hands-free voice mode, so you know what to go try today.","aiDescription":"<p><strong>Introduction and Guest Overview:</strong> The session opens with greetings as the hosts confirm they are live (<a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>). Evan Boyle from the GitHub Copilot CLI team is introduced as a guest and described as both an engineering manager and hands-on developer passionate about developer tools (<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>). He explains that his team builds several components of Copilot including the core agent loop, SDK, CLI, and the newly released GitHub app, which were also highlighted during a keynote presentation. The atmosphere is casual and collaborative, as they recall working late nights preparing demos at GitHub HQ prior to the event (<a class=\"medius-aisummary\" data-starttime=\"00:01:08\">00:01:08</a>).</p>\n\n<p><strong>Copilot CLI Demonstration:</strong> Evan transitions into demonstrating Copilot CLI, prompted by a conversation with someone who asked what it was (<a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>). He walks through a live example using a GitHub repository that was built interactively with the community during a previous session (<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>). The audience submitted about sixty issues, and the Copilot CLI was used to analyze them and select one for implementation. That issue resulted in an Electron app featuring an animated Octocat called “goblin mode” that reacts to Git operations with emojis and sounds (<a class=\"medius-aisummary\" data-starttime=\"00:02:57\">00:02:57</a>). He showcases how the CLI’s plan mode automates project building while interacting with user feedback live, demonstrating that the tool can rapidly generate creative projects based on open-source collaboration.</p>\n\n<p><strong>Working in Terminals and Session Management:</strong> After the demo, they discuss the appeal of terminal-based development and explain why the CLI runs there instead of inside an editor (<a class=\"medius-aisummary\" data-starttime=\"00:04:43\">00:04:43</a>). Evan emphasizes the focus and simplicity of the terminal environment — no distractions, real-time logging, and the ability to feel “like a hacker.” He explains how Copilot CLI now supports multiple parallel sessions through commands like <code>slash new</code> and <code>slash sessions</code> (<a class=\"medius-aisummary\" data-starttime=\"00:05:27\">00:05:27</a>), enabling users to manage different tasks simultaneously. The functionality is demonstrated as he closes issues in one session while researching others in a backgrounded session (<a class=\"medius-aisummary\" data-starttime=\"00:06:01\">00:06:01</a>). This multi-session runtime models efficiency for developers managing complex workflows directly in the terminal.</p>\n\n<p><strong>Team Velocity and Slash Commands:</strong> The discussion then shifts to GitHub’s rapid internal development culture (<a class=\"medius-aisummary\" data-starttime=\"00:09:07\">00:09:07</a>). Evan highlights that the team ships thousands of PRs monthly across multiple components, which means the codebase evolves quickly. He then introduces favorite Copilot CLI features that rely on slash commands, like <code>slash review</code> for multi-model code reviews using different AI engines (<a class=\"medius-aisummary\" data-starttime=\"00:10:14\">00:10:14</a>). The CLI showcases its ability to intelligently prompt users when ambiguity exists, performing “elicitation” to clarify intent before running reviews (<a class=\"medius-aisummary\" data-starttime=\"00:11:14\">00:11:14</a>). The conversation explores how developers can customize verbosity levels for AI responses, balancing detailed output versus concise summaries, and even modify “copilot instructions” using voice input for personal workflow configurations (<a class=\"medius-aisummary\" data-starttime=\"00:13:14\">00:13:14</a>).</p>\n\n<p><strong>Voice Mode and Workflow Personalization:</strong> Evan demonstrates the CLI’s voice mode capabilities, enabled through <code>slash voice</code> and local language models (<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>). He explains that these models, provided locally and running offline, allow real-time voice interactions without cloud dependency. The hosts discuss user comfort levels with speaking aloud to AI agents and reflect on how voice input activates different cognitive processes than typing (<a class=\"medius-aisummary\" data-starttime=\"00:15:03\">00:15:03</a>). Evan mentions learning from developer Scott Hanselman, who uses voice mode extensively. He shares that he personally uses a remote control to pace around his office and dictate tasks — illustrating an evolving, natural workflow between humans and AI agents that merges technology with typical thought processes (<a class=\"medius-aisummary\" data-starttime=\"00:15:57\">00:15:57</a>).</p>\n\n<p><strong>Closing and Next Steps:</strong> As the session concludes, Evan announces upcoming talks, including one later that day on the GitHub app with Mario and another recorded deep dive with Cassidy on Copilot CLI (<a class=\"medius-aisummary\" data-starttime=\"00:16:25\">00:16:25</a>). He encourages viewers to experiment with Copilot CLI firsthand, emphasizing how surprisingly smooth and productive it feels despite any initial skepticism (<a class=\"medius-aisummary\" data-starttime=\"00:16:53\">00:16:53</a>). The hosts wrap up with thanks and final encouragement to try and explore these emergent AI-driven development tools. The broadcast closes warmly, pointing developers toward hands-on adoption as the next step (<a class=\"medius-aisummary\" data-starttime=\"00:17:22\">00:17:22</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/609cb95c-6575-4206-9586-c26853e87efd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/609cb95c-6575-4206-9586-c26853e87efd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/609cb95c-6575-4206-9586-c26853e87efd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/609cb95c-6575-4206-9586-c26853e87efd","speakerIds":["1768245294609build26-1776135903186001Orl7","1768245294609build26-1776347378057001Rvny"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/609cb95c-6575-4206-9586-c26853e87efd","ogTitle":"What's new in GitHub Copilot CLI?","ogDescription":"This session walks through what's new, including a redesigned terminal interface, Rubber Duck for second opinions, recurring prompts with /every, and hands-free voice mode, so you know what to go try today.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776135903186001Orl7","RegistrantKey":"1768245294609build26-1776135903186001Orl7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/GHCP/CLI","backpackLink":null,"label":"Get started with GitHub Copilot CLI","category":"Web link"},{"link":"https://aka.ms/GHCP/CLI-bestpractices","backpackLink":null,"label":"Learn GitHub Copilot CLI best practices","category":"Web link"},{"link":"https://aka.ms/GHCP/CLI-Learn","backpackLink":null,"label":"New to the GitHub Copilot CLI? Watch the GitHub Copilot CLI for Beginners series","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d5062c96-7ebd-490f-bc4e-5bb561627f97","startDateTime":"2026-06-02T23:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Charles Feddersen, Sachin Patney, Nikisha Reyes-Grange","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779991829481001vfVg","langLocale":"en-US","sessionId":"1779991829481001vfVg","sessionInstanceId":"1779991829481001vfVg","sessionCode":"LIVE143","sessionCodeNormalized":"EN-US-LIVE143","title":"What’s New in Azure Data: HorizonDB and Rayfin","sortTitle":"what’s new in azure data: horizondb and rayfin","description":"Get to know two exciting data innovations. Azure HorizonDB rethinks how PostgreSQL databases are built and operated, focusing on a modern, scalable foundation designed for cloud‑native applications. Rayfin makes it easier to build and run data applications in Microsoft Fabric, bringing data, logic, and AI closer together so developers can scale faster without stitching backend services. Learn the motivation behind each, the core concepts to understand, and how they shape the future of application and data development.","aiDescription":"<p><strong>Introduction and Speaker Overview:</strong> The video begins with Nikisha Reyes Grange greeting attendees and wishing them a happy Build conference <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. She introduces herself as Senior Director with Azure Data and AI and welcomes her colleagues Charles Federsen and Sachin Patne <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>. Charles, the Director of Program Management for Postgres and MySQL on Azure, and Sachin, the General Manager of App Development for Azure Data, join to share key data announcements made at Build. The main agenda includes two major topics: Azure Horizon DB and Raefen SDK <a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a>.</p>\n\n<p><strong>Azure Horizon DB Announcement:</strong> Charles explains that Azure Horizon DB is a newly announced Postgres-based database service designed for developers and enterprises <a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>. It entered public preview at the time of the Build event <a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>. Horizon DB aims to address both app development and enterprise-scale needs by integrating AI and developer-centric tooling. It supports enhanced performance, enterprise-grade security, and AI-driven features, while allowing seamless development in VS Code. Charles highlights new developer enhancements like AI pipelines for asynchronous workflows <a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a> and improved storage architecture offering higher scalability and availability compared to standard Postgres services <a class=\"medius-aisummary\" data-starttime=\"00:03:11\">00:03:11</a>. He notes that developers benefit from full Postgres compatibility.</p>\n\n<p><strong>Introduction to Raefen SDK and Its Purpose:</strong> Sachin then introduces Raefen, describing it as an open-source SDK and CLI for defining backend services entirely in code <a class=\"medius-aisummary\" data-starttime=\"00:03:49\">00:03:49</a>. Raefen integrates with Azure Fabric, automatically translating databases, functions, and authentication components into governed, scalable Fabric services during deployment <a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>. It was created to bridge the gap between rapid app prototyping and enterprise-grade production deployment <a class=\"medius-aisummary\" data-starttime=\"00:05:02\">00:05:02</a>. Raefen allows developers to focus on business logic while ensuring governance, compliance, and scalability. Sachin emphasizes that it works seamlessly with other app-building platforms, such as GitHub Copilot and Replit, empowering developers to deploy managed backends directly into their Fabric environments <a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>.</p>\n\n<p><strong>Raefen’s Role within Microsoft Fabric and Open Source Vision:</strong> The discussion moves to why Raefen is deeply integrated with Fabric <a class=\"medius-aisummary\" data-starttime=\"00:06:25\">00:06:25</a>. Since Fabric is a SaaS platform with zero infrastructure management, Raefen enables organizations to build and host operational and even customer-facing applications without complex data movement. Application data stays within Fabric for instant analytics and governance <a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>. Sachin also notes that parts of Raefen’s runtime will be open-sourced, offering flexibility to self-host applications or run outside Fabric if preferred <a class=\"medius-aisummary\" data-starttime=\"00:08:10\">00:08:10</a>. This approach allows the community to extend Raefen across various infrastructures, making it suitable for both light and enterprise-scale applications <a class=\"medius-aisummary\" data-starttime=\"00:08:21\">00:08:21</a>.</p>\n\n<p><strong>Postgres Community Commitment:</strong> Charles highlights Microsoft’s significant role in supporting the open-source Postgres community <a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>. Microsoft is among the largest contributors to the upstream Postgres project, modifying roughly 8% of the codebase in the upcoming version 19 <a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>. This contribution ensures that insights gained from managing cloud-scale Postgres instances benefit the entire ecosystem. Microsoft also organizes the largest virtual Postgres event globally, maintains a podcast, and participates actively in community conferences <a class=\"medius-aisummary\" data-starttime=\"00:09:56\">00:09:56</a>. The company’s vision is to ensure that enterprise customers running mission-critical workloads receive robust, secure, and well-supported Postgres environments <a class=\"medius-aisummary\" data-starttime=\"00:10:08\">00:10:08</a>.</p>\n\n<p><strong>AI and Hybrid Search Capabilities in Horizon DB:</strong> Returning to Horizon DB, Charles elaborates on its AI features <a class=\"medius-aisummary\" data-starttime=\"00:10:46\">00:10:46</a>. Building on Postgres’s PG vector extension, Horizon introduces AI functions that interact directly with Microsoft Foundry models from SQL <a class=\"medius-aisummary\" data-starttime=\"00:11:30\">00:11:30</a>. Developers can instantly use pre-registered embedding and re-ranking models without complex setup. A new hybrid search function fuses full-text and vector searches, allowing highly relevant and high-performance retrieval across enterprise data <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:24\">00:12:24</a>. The system supports multiple indexing options (HNSW, IV flat) and even graph-based data relationships visualized through VS Code extensions <a class=\"medius-aisummary\" data-starttime=\"00:13:01\">00:13:01</a>. The session ends with Nikisha thanking the speakers and wishing viewers a happy Build <a class=\"medius-aisummary\" data-starttime=\"00:13:09\">00:13:09</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"23:20 - 23:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e6e02047-ba96-4845-9f56-11be582215ec","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e6e02047-ba96-4845-9f56-11be582215ec","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e6e02047-ba96-4845-9f56-11be582215ec","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e6e02047-ba96-4845-9f56-11be582215ec","speakerIds":["1768245294609build26-1765362540584001w9jd","1768245294609build26-1774284996963001pGpM","1768245294609build26-1756140167720001oG8r"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.459Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e6e02047-ba96-4845-9f56-11be582215ec","ogTitle":"What’s New in Azure Data: HorizonDB and Rayfin","ogDescription":"Get to know two exciting data innovations. Azure HorizonDB rethinks how PostgreSQL databases are built and operated, focusing on a modern, scalable foundation designed for cloud‑native applications. Rayfin makes it easier to build and run data applications in Microsoft Fabric, bringing data, logic, and AI closer together so developers can scale faster without stitching backend services. Learn the motivation behind each, the core concepts to understand, and how they shape the future of application and data development.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756140167720001oG8r","RegistrantKey":"1768245294609build26-1756140167720001oG8r"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362540584001w9jd","RegistrantKey":"1768245294609build26-1765362540584001w9jd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996963001pGpM","RegistrantKey":"1768245294609build26-1774284996963001pGpM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"71a9324f-b774-4095-a16d-328db6918f2d","startDateTime":"2026-06-02T22:35:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Guthrie, Seth Juarez","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779512743066001aiqi","langLocale":"en-US","sessionId":"1779512743066001aiqi","sessionInstanceId":"1779512743066001aiqi","sessionCode":"LIVE109","sessionCodeNormalized":"EN-US-LIVE109","title":"What’s real, ready, and next for developers with Scott Guthrie","sortTitle":"what’s real, ready, and next for developers with scott guthrie","description":"There’s a lot of noise about AI. In this quick-paced fireside chat, get the builder’s perspective on what actually matters for developers. Drawing from how Microsoft runs a complete system from silicon to software at global scale, Scott breaks down the evolution across AI-ready infrastructure, context layers for agents, and what AI-assisted modernization means for developers.","aiDescription":"<p><strong>Introduction and Context:</strong> The video begins with Seth Juarez welcoming viewers back to Build and introducing his guest, Scott Guthrie (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a>). Guthrie explains his role at Microsoft as leading Azure and the broader cloud infrastructure (<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>). He emphasizes that Azure serves as the foundation for everything Microsoft builds, from infrastructure and data platforms to higher-level services. At the top of his priorities are expanding global capacity, reducing AI costs through custom silicon (Azure Maia and Azure Cobalt), and maintaining reliability and high performance for AI workloads while enabling tools like GitHub Copilot and Microsoft 365 Copilot (<a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>).</p>\n\n<p><strong>Infrastructure Scale and Sustainability:</strong> The discussion moves into the sheer scale of Azure’s global operations (<a class=\"medius-aisummary\" data-starttime=\"00:01:37\">00:01:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:49\">00:02:49</a>). Guthrie shares that Microsoft recently added over a gigawatt of datacenter capacity within 90 days, equivalent to powering a city the size of Seattle (<a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>). He stresses Microsoft’s commitment to responsible growth, noting its 0-water-waste design and efforts to support local communities through energy stewardship and engagement. Operating in over 80 locations worldwide requires precise planning and execution, backed by skilled tradespeople, strong safety practices, and innovative engineering to ensure both safety and reliability for global customers (<a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:56\">00:03:56</a>).</p>\n\n<p><strong>Engineering Excellence and Technical Depth:</strong> Guthrie elaborates on the complexity of scaling Azure’s software and hardware architecture as capacity doubles every two years (<a class=\"medius-aisummary\" data-starttime=\"00:04:02\">00:04:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:47\">00:05:47</a>). He highlights supply chain constraints in memory and storage and explains how continuous software innovation is required to match growth. To support AI and agentic workloads, Microsoft redesigned core systems such as Cosmos DB to handle global scale by distributing databases regionally with eventual consistency, ensuring seamless replication and fault tolerance (<a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>). He uses examples like ChatGPT and Copilot to illustrate challenges like chat history management, attachment storage in blob systems, and schema evolution—all done live without downtime.</p>\n\n<p><strong>Azure Data Services and Innovations:</strong> The conversation transitions into Azure’s data ecosystem, covering SQL, NoSQL, and Postgres-compatible systems (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>). Guthrie categorizes offerings into SQL services such as Hyperscale for SaaS workloads, Cosmos DB supporting large-scale AI solutions, and Horizon DB—a new Postgres service designed for horizontal scale with built-in AI and enhanced elasticity (<a class=\"medius-aisummary\" data-starttime=\"00:08:23\">00:08:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:53\">00:08:53</a>). He introduces Microsoft Fabric, the unified analytics platform with “OneLake,” which integrates data from Databricks, Snowflake, and Power BI models under a single system without the need for data transfer. Fabric IQ—now generally available—connects semantic data directly to Copilot in Microsoft 365 for more intelligent organizational insights (<a class=\"medius-aisummary\" data-starttime=\"00:09:40\">00:09:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>).</p>\n\n<p><strong>Scale, AI Pressure, and Global Replication:</strong> Seth asks about Azure’s massive data scale, with Guthrie revealing that Microsoft now adds exabytes of storage per week—an exponential evolution from traditional petabyte capacities (<a class=\"medius-aisummary\" data-starttime=\"00:10:26\">00:10:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:03\">00:11:03</a>). As AI agents and automated systems introduce millions of requests per second, the architecture must evolve beyond conventional relational designs to globally replicated systems capable of ultra-low latency. These advances ensure apps remain performant under unprecedented load while allowing hybrid use of relational and distributed no-SQL models to balance consistency and scalability (<a class=\"medius-aisummary\" data-starttime=\"00:11:36\">00:11:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:59\">00:11:59</a>).</p>\n\n<p><strong>AI Transformation and Future Outlook:</strong> In closing, Guthrie reflects on how generative and agentic AI will reshape the software landscape, drawing comparisons to prior waves like mobile and web revolutions (<a class=\"medius-aisummary\" data-starttime=\"00:12:05\">00:12:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:15\">00:13:15</a>). He stresses that AI’s success depends on clean, scalable data, enabled by technologies like Fabric IQ and efficient infrastructure. Advancements at every layer—from VM startup times and ephemeral compute environments to Cobalt processors and improved memory subsystems—are designed for agile, short-lived workloads characteristic of intelligent agents (<a class=\"medius-aisummary\" data-starttime=\"00:13:37\">00:13:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:15\">00:14:15</a>). The video concludes as Seth thanks Guthrie, who confirms that the upcoming AI-driven transformation will be vast and exhilarating for developers and technologists alike (<a class=\"medius-aisummary\" data-starttime=\"00:14:16\">00:14:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:22\">00:14:22</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"22:35 - 22:50","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:50:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/66e36077-6ee0-4e48-a7e4-c206c235379b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/66e36077-6ee0-4e48-a7e4-c206c235379b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/66e36077-6ee0-4e48-a7e4-c206c235379b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/66e36077-6ee0-4e48-a7e4-c206c235379b","speakerIds":["1768245294609build26-1748558014623001cxi9","1768245294609build26-1751379013075001HHpX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.237Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/66e36077-6ee0-4e48-a7e4-c206c235379b","ogTitle":"What’s real, ready, and next for developers with Scott Guthrie","ogDescription":"There’s a lot of noise about AI. In this quick-paced fireside chat, get the builder’s perspective on what actually matters for developers. Drawing from how Microsoft runs a complete system from silicon to software at global scale, Scott breaks down the evolution across AI-ready infrastructure, context layers for agents, and what AI-assisted modernization means for developers.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1748558014623001cxi9","RegistrantKey":"1768245294609build26-1748558014623001cxi9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751379013075001HHpX","RegistrantKey":"1768245294609build26-1751379013075001HHpX"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d21d14c7-439d-4bd9-86b0-1f93273a9445","startDateTime":"2026-06-03T19:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Justin Chen, Burke Holland","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780008268215001UNRW","langLocale":"en-US","sessionId":"1780008268215001UNRW","sessionInstanceId":"1780008268215001UNRW","sessionCode":"LIVE154","sessionCodeNormalized":"EN-US-LIVE154","title":"Your Browser Lives in VS Code Now","sortTitle":"your browser lives in vs code now","description":"The integrated browser is not just a preview pane. Agents can share tabs as context, read page content, interact with elements, run Playwright scripts, validate changes in real time, and debug with breakpoints without ever leaving the editor. This session shows what that actually looks like in a real development workflow.","aiDescription":"<p><strong>Session Introduction:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>, the speaker welcomes both online and in-person audiences at Build, expressing appreciation for attendees who have stuck around throughout the event. He introduces Justin from the Visual Studio Code (VS Code) team, setting the stage for a discussion focused on new features developed for the popular editor. The atmosphere feels casual and friendly, with light banter between the host and audience before diving into the technical content.</p>\n\n<p><strong>Overview of the Integrated Browser:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:09\">00:01:09</a>, Justin introduces the main topic—the integrated browser feature inside VS Code. He explains that the goal was to improve local development by providing in-editor browsing capabilities, removing the need to switch to external browsers. This tool can host local websites, log in to external sites, and enable developers to preview projects directly within VS Code. Built from the simple browser extension about a year ago, the integrated browser now allows seamless workflow by letting developers test and interact with their projects in one environment.</p>\n\n<p><strong>Feature Demonstration:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:01:27\">00:01:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:05\">00:03:05</a>, the demonstration begins as Justin shares his screen and opens the integrated browser through the command palette. He shows how it can open websites like GitHub and locally hosted apps, including a personal Pokémon-themed application he developed. Humor and casual commentary punctuate the demo as the host jokes about Pokémon references while Justin emphasizes the functionality of browsing and testing sites directly in the editor. This section highlights the integrated browser’s flexibility, user authentication, and instant project launch features.</p>\n\n<p><strong>Deep Dive into Development Tools:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:09\">00:05:09</a>, Justin explores advanced capabilities like adding DOM elements into a chat interface and capturing screenshots with contextual metadata (such as element path, URL, and outer HTML). He explains that users can customize included data—for instance, disabling CSS if unnecessary. The integrated browser also embeds full Chrome DevTools, allowing developers to inspect elements, execute JavaScript in the console, and emulate mobile devices. The newly added emulation toolbar provides preset screen dimensions and device profiles, making it easier to test mobile responsiveness within VS Code.</p>\n\n<p><strong>Integration with AI Agent and Playwright:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:05:51\">00:05:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:09\">00:08:09</a>, the speakers demonstrate how VS Code’s integrated agent can control the browser without additional tools like Playwright or external integration. By clicking \"share with agent,\" a glowing border indicates the browser is under agent control, enabling automated navigation and testing. The team discusses challenges such as token efficiency in AI use, advising precise element selection to reduce resource costs. The agent executes tasks like navigating tabs and running end-to-end tests, though occasional debugging is required when the agent struggles to locate specific page elements.</p>\n\n<p><strong>Playwright Integration and Conclusion:</strong> Finally, at <a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>, Justin elaborates on the Playwright integration, showing how tests can automatically run across different viewports and device sizes using built-in browser emulation. He emphasizes that developers do not need to install anything extra—this capability is included for free with VS Code. The session ends with lighthearted banter and thanks as the host wraps up, celebrating the integrated browser as a powerful, accessible tool that merges coding, testing, and automation directly in one editor.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":2,"TimeSlot":"19:45 - 19:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:55:00Z","durationInMinutes":10,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/cc424a49-15c3-479e-99d9-830cac9021ca","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/cc424a49-15c3-479e-99d9-830cac9021ca","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/cc424a49-15c3-479e-99d9-830cac9021ca","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/cc424a49-15c3-479e-99d9-830cac9021ca","speakerIds":["1768245294609build26-1774285001041001p103","1768245294609build26-1776347378057001Rvny"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.462Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/cc424a49-15c3-479e-99d9-830cac9021ca","ogTitle":"Your Browser Lives in VS Code Now","ogDescription":"The integrated browser is not just a preview pane. Agents can share tabs as context, read page content, interact with elements, run Playwright scripts, validate changes in real time, and debug with breakpoints without ever leaving the editor. This session shows what that actually looks like in a real development workflow.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Broadcast Stage","logicalValue":"Broadcast Stage"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Build Broadcast Stage","logicalValue":"Gateway Pavilion, Level 1, Build Broadcast Stage"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285001041001p103","RegistrantKey":"1768245294609build26-1774285001041001p103"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/IntBrowser","backpackLink":null,"label":"Try the integrated browser in VS Code","category":"Web link"},{"link":"https://aka.ms/VSCode/Learn","backpackLink":null,"label":"Grow your VS Code and GitHub Copilot skills with free, hands-on course","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"bff83108-9508-436c-b710-a6199c4cb42a","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"swyx (Shawn Wang)","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701128001r9yb","langLocale":"en-US","sessionId":"1774018701128001r9yb","sessionInstanceId":"1774018701128001r9yb","sessionCode":"BRK244","sessionCodeNormalized":"EN-US-BRK244","title":"Agent supervision is the new senior engineering skill","sortTitle":"agent supervision is the new senior engineering skill","description":"As AI agents start shipping code, the bottleneck becomes trust. This talk argues “agent supervision” is the new senior skill: scoping agent work, setting constraints, designing checkpoints, and reviewing outputs for correctness and security. Learn practical patterns to catch silent failures and build workflows where humans stay accountable and agents stay fast.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Industry Context:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>, Swix opens his talk with observations on current developments in the AI engineering landscape, humorously referencing the launch and rebranding of the GitHub Copilot app as emblematic of a wider trend in “agent command centers.” He notes that competing tools like Cursor, Windsurf, and Codex have adopted similar interfaces, converging toward a standard form factor for managing AI agents. This convergence, described around <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>, mirrors natural evolution in technology design, which he compares to “everything becoming a crab.” The speaker sets the stage for his broader argument that this evolution signals an agreed direction for how developers will interact with intelligent systems going forward.</p>\n\n<p><strong>Agent-Centric Coding and Community Dynamics:</strong> By <a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>, Swix shifts into examining how agent tools are reshaping developer workflows, referencing Cognition’s recent product that introduces a local-to-cloud handoff for managing multiple agents. He reflects on his own chaotic development environment and how such tools aim to simplify complex AI orchestration. Then, at <a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:07\">00:03:07</a>, the focus turns to the explosive growth of the global AI engineering community across cities like Paris, London, and Singapore. He highlights six recurring themes shaping this growth, including the shift from “model worship” to combining models with harnesses, resulting in a more developer-centric environment. Swix introduces the concept of “format fear of missing agent time,” describing the pressure engineers feel to constantly run or update agents, capturing the intensity and transformative nature of the field.</p>\n\n<p><strong>Evolution of Coding Agents and Development Cycles:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:06:17\">00:06:17</a>, Swix discusses how longer autonomous task capabilities in models have created new UX requirements and workflows. GitHub and others are seeing exponential increases in AI-generated code, with cloud-based code commits rising from roughly 4–5% early in the year to an expected 50% by year-end. He references a 14,000× growth in commit activity reported by GitHub’s CEO (<a class=\"medius-aisummary\" data-starttime=\"00:07:14\">00:07:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>). These developments, he argues, are spreading beyond coding into general computer use, browsing, and design. As seen in projects merging Codex into ChatGPT (<a class=\"medius-aisummary\" data-starttime=\"00:16:17\">00:16:17</a>), Swix concludes that the paradigm of multi-agent orchestration is expanding across all domains of digital work.</p>\n\n<p><strong>Software Development Life Cycle and the Dark Factory:</strong> Beginning near <a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>, the conversation moves to redefining the software development life cycle (SDLC) in the age of autonomous agents. Traditional pull requests could evolve into “prompt requests,” allowing maintainers to adapt intent instead of reviewing human code. Swix details concepts like the “dark factory,” in which AI agents design, code, and review software with minimal to zero human oversight (<a class=\"medius-aisummary\" data-starttime=\"00:10:15\">00:10:15</a>). He answers audience questions about hardening validation and change management, emphasizing a layered control model using specifications, strong test suites, progressive rollouts, and online evaluation loops. Discussion of Ryan Le Papolo’s “harness engineering” work (<a class=\"medius-aisummary\" data-starttime=\"00:13:45\">00:13:45</a>) illustrates the scale of these autonomous systems—shipping billions of tokens daily without human review—while stressing the need for better communication bandwidth between humans and agents to maintain alignment.</p>\n\n<p><strong>Compute Infrastructure and Emerging Research Directions:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a> onward, Swix delves into the computational layer enabling large-scale inference. He contrasts historical GPU-heavy workloads (8:1 ratios) with new CPU-heavy demands (approaching parity), forecasting a CPU shortage as AI tasks scale. Examples such as Talos’ 17,000 tokens per second demo foreshadow breakthroughs leading to 300,000 tokens per second inference speeds (<a class=\"medius-aisummary\" data-starttime=\"00:23:22\">00:23:22</a>). The speaker situates this within a longer-term hardware and model evolution, where open-weight models and optimized inference extend chip life cycles from three to eight years (<a class=\"medius-aisummary\" data-starttime=\"00:28:22\">00:28:22</a>). He then transitions to upcoming AI research priorities (<a class=\"medius-aisummary\" data-starttime=\"00:30:29\">00:30:29</a>): real-time responsiveness, memory and continual learning, and interactive micro-batching aimed at human-machine “mind melds.” This vision includes video generation driven by orchestrating agents and the rise of world models capable of predicting user preferences and intentions.</p>\n\n<p><strong>Conclusion and Reflections on the AI Engineering Frontier:</strong> In his closing remarks at <a class=\"medius-aisummary\" data-starttime=\"00:34:57\">00:34:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:35:34\">00:35:34</a>, Swix introduces the “agent work paradox”—even as agents take over more tasks, engineers are working harder than ever to manage them. He normalizes this tension as a sign of progress, framing AI engineering as “the last job” that will automate all others. Despite the stress and accelerated pace, he expresses optimism: developers today are living the future first, laying groundwork for automation across knowledge work, creativity, and infrastructure. The presentation concludes on a motivational note, encouraging engineers to embrace their role at the forefront of this transformative era in human-machine collaboration.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c8e844ff-d07c-427f-9168-2a7d84ac4acf","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c8e844ff-d07c-427f-9168-2a7d84ac4acf","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c8e844ff-d07c-427f-9168-2a7d84ac4acf","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c8e844ff-d07c-427f-9168-2a7d84ac4acf","speakerIds":["1768245294609build26-1775516734838001pZmI"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.456Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c8e844ff-d07c-427f-9168-2a7d84ac4acf","ogTitle":"Agent supervision is the new senior engineering skill","ogDescription":"As AI agents start shipping code, the bottleneck becomes trust. This talk argues “agent supervision” is the new senior skill: scoping agent work, setting constraints, designing checkpoints, and reviewing outputs for correctness and security. Learn practical patterns to catch silent failures and build workflows where humans stay accountable and agents stay fast.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775516734838001pZmI","RegistrantKey":"1768245294609build26-1775516734838001pZmI"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"OSS CI/CD Libraries","logicalValue":"OSS CI/CD Libraries"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"}]},{"@search.score":1.0,"documentId":"85d35b6d-d0a8-429c-96d4-e6312c385837","startDateTime":"2026-06-03T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"David Fowler, Maddy Montaquila, PJ Meyer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709091001rtSa","langLocale":"en-US","sessionId":"1774018709091001rtSa","sessionInstanceId":"1774018709091001rtSa","sessionCode":"BRK205","sessionCodeNormalized":"EN-US-BRK205","title":"Aspire for agents: Transform how you build and deploy distributed apps","sortTitle":"aspire for agents: transform how you build and deploy distributed apps","description":"Aspire is an open-source cloud and agent-ready developer toolchain transforming how you build, run, and deploy observable, production-ready distributed apps. Its code centric approach streamlines orchestration of your apps and agents, provides observability for systems in any programming language, and supports deployment locally or to your cloud of choice. Come meet the new Aspire.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Purpose of the Session:</strong> The session opens with a lively introduction (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:18\">00:00:18</a>), where the presenters welcome attendees to the “Aspire for Agents” demo focused on transforming how distributed apps are built and deployed. After ensuring everyone can hear and has access to headsets, they clarify what constitutes a distributed application—emphasizing that even a simple website with a database qualifies. The presenters introduce themselves and outline the session’s structure: a fast-paced 45-minute demonstration primarily focused on live code, supported by easily shareable slides for reference. They encourage audience participation both during and after the session at their booth for further discussion.</p>\n\n<p><strong>What Aspire Is and Why It Matters:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:01:11\">00:01:11</a>, the speakers define Aspire as a tool designed to simplify the process of composing, debugging, and deploying distributed systems. They highlight the common pain point of moving from “it works on my machine” to containerized, cloud, or Kubernetes environments where behavior shifts dramatically. Aspire’s one-line summary, introduced at <a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>, describes it as an “agent-ready, code-first tool to compose, debug, and deploy any distributed app.” They explain Aspire’s modular nature—comprising a CLI, an app host for defining orchestration logic, preconfigured integrations for common technologies, and a dashboard for visibility. By focusing on the local developer experience and seamlessly extending to deployment, Aspire aims to eliminate configuration drift between development and production environments.</p>\n\n<p><strong>The Aspire Dashboard and Integration with Agents:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:06:40\">00:06:40</a> onward, the presentation explores the Aspire dashboard and how it streamlines development workflows. Using open telemetry (“otel”) support, Aspire enables real-time insights into logs, traces, and performance metrics across complex systems. The presenters demonstrate how the dashboard can serve as a stand-alone telemetry viewer or as an onboarding tool for new developers, acting as a “Trojan horse” for broader Aspire adoption. They emphasize the product’s evolution to optimize both human and AI coding agent workflows, illustrating how agents can query logs or performance data directly through Aspire’s CLI commands. The design ensures parity between what human developers and coding agents can access, making debugging and collaboration easier in modern multi-agent coding environments.</p>\n\n<p><strong>Live Demo: Building and Managing a Distributed App:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:10:18\">00:10:18</a>, the speakers begin a live demonstration of Aspire’s CLI capabilities. They show installation via Winget or Homebrew and proceed to use `Aspire start` to launch their demo application. This “calendar booking” app includes a web front end, an API layer, and an autonomous background agent. As they run Aspire, it automatically detects dependencies, sets environment variables, allocates ports to avoid conflicts, and starts related processes such as Postgres containers. Through the dashboard, they execute administrative commands like clearing or generating calendars directly within the interface. These dynamic commands highlight how Aspire’s app host configuration can define both runtime resources and administrative controls without manual shell commands or external scripts. The presenters then show how all these components—API, database, and agent—communicate securely while maintaining developer visibility through telemetry and logs.</p>\n\n<p><strong>Advanced Configuration, Deployment, and Cloud Integration:</strong> The latter part of the demo (<a class=\"medius-aisummary\" data-starttime=\"00:28:00\">00:28:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:44:00\">00:44:00</a>) dives deeper into customization, environment variables, and deployment automation. Using “app host” definitions in C# or TypeScript, Aspire enables developers to define relationships among services—like web apps referencing API endpoints—without hardcoding ports or URLs. The presenters demonstrate adding support for Microsoft Foundry (a hosted agent platform) and illustrate automatic Azure integration, where Aspire provisions cloud resources, sets permissions, and injects secure configuration values. Aspire’s model ensures sensitive credentials are stored safely in user profiles rather than `.env` files to prevent leaks. They also showcase deployment using `Aspire deploy`, which translates the defined model into container app environments and compatible infrastructure-as-code templates like Bicep or YAML. This workflow allows developers to seamlessly move from local development to production-grade cloud orchestration in a consistent, repeatable way.</p>\n\n<p><strong>Conclusion, Resources, and Community:</strong> As the session concludes at <a class=\"medius-aisummary\" data-starttime=\"00:45:00\">00:45:00</a>, the presenters direct attendees to the latest updates at Aspire.dev and invite them to join the community Discord and weekly livestreams. They reaffirm Aspire’s value as a tool that simplifies distributed development, empowers AI agents, and standardizes the path from prototype to deployed system. The talk ends on a personal note, with mentions of real engineers becoming so accustomed to Aspire that they refuse to work without it. The team encourages everyone to experiment with the provided sample code, share feedback, and explore how Aspire can enhance productivity in both human- and agent-driven development environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:45 - 22:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","speakerIds":["1768245294609build26-1774982456082001foFO","1768245294609build26-1774982579205001NKRR","1768245294609build26-1774284989075001pX2D"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.788Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e19a5e82-a2b2-4e2f-acdf-804ccc5a0f5b","ogTitle":"Aspire for agents: Transform how you build and deploy distributed apps","ogDescription":"Aspire is an open-source cloud and agent-ready developer toolchain transforming how you build, run, and deploy observable, production-ready distributed apps. Its code centric approach streamlines orchestration of your apps and agents, provides observability for systems in any programming language, and supports deployment locally or to your cloud of choice. Come meet the new Aspire.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774284989075001pX2D","RegistrantKey":"1768245294609build26-1774284989075001pX2D"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774982456082001foFO","RegistrantKey":"1768245294609build26-1774982456082001foFO"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774982579205001NKRR","RegistrantKey":"1768245294609build26-1774982579205001NKRR"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK205","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"},{"displayValue":"Aspire","logicalValue":"Aspire"}]},{"@search.score":1.0,"documentId":"5ca207c1-819d-484e-b82c-81ccb929b8af","startDateTime":"2026-06-03T20:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dave Burnison, Dan Hellem, Lan Kaim","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708576001rogc","langLocale":"en-US","sessionId":"1774018708576001rogc","sessionInstanceId":"1774018708576001rogc","sessionCode":"BRK202","sessionCodeNormalized":"EN-US-BRK202","title":"Azure DevOps meets GitHub, the path to AI powered SDLC","sortTitle":"azure devops meets github, the path to ai powered sdlc","description":"Azure DevOps and GitHub are better together—and the integration keeps getting smarter. In this demo-heavy session, you’ll see how hybrid patterns that connect GitHub with Azure Boards and Azure Pipelines enable Agentic DevOps. See the all of the newest AI-powered capabilities in Azure DevOps. Plus, hear how Microsoft's engineering teams adopted this approach and what they gained.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Goals:</strong> The webinar begins with greetings from <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a> by Dave Burnison, Senior DevOps Advocate at GitHub, joined by Dan Hellem, Project Manager on Azure DevOps. They outline the session’s focus on \"Azure DevOps and GitHub: Your Path to Agentic AI\" <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>. The goal is to demonstrate how integrating GitHub with Azure DevOps can empower development teams with agentic AI capabilities. Dave explains that while GitHub and Azure DevOps have long operated as distinct ecosystems, the convergence of tools now enables improved workflows and AI-driven productivity whether teams are startups or large enterprises <a class=\"medius-aisummary\" data-starttime=\"00:01:31\">00:01:31</a>. He emphasizes leveraging GitHub Copilot’s intelligence while maintaining existing investments in Azure boards, pipelines, and test plans <a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>.</p>\n\n<p><strong>Practical Demonstrations: Integrating Copilot and Azure DevOps:</strong> Transitioning to demos <a class=\"medius-aisummary\" data-starttime=\"00:03:09\">00:03:09</a>, Dave presents how project managers can harness the GitHub Copilot app within Azure boards using minimal setup. He configures the Azure DevOps remote MCP server through settings, linking organizational URLs and selecting tools such as work items, pipelines, and test plans while omitting GitHub repos <a class=\"medius-aisummary\" data-starttime=\"00:04:19\">00:04:19</a>. Demonstrations reveal querying active work items, viewing linked user stories, and integrating with Azure pipelines to review run statuses <a class=\"medius-aisummary\" data-starttime=\"00:06:27\">00:06:27</a>. He showcases how the Copilot app refines documentation standards by converting Azure Wiki best‑practice pages into reusable Copilot instructions and agent files in markdown <a class=\"medius-aisummary\" data-starttime=\"00:08:02\">00:08:02</a>. This workflow enables Copilot to verify adherence of user stories to established practices, suggest improvements, and even automate refinement tasks directly in Azure boards.</p>\n\n<p><strong>Enhancing Code and Security Collaboration:</strong> Continuing the demo sequence, Dave illustrates how Copilot agents update backlog items, enhance acceptance criteria, and handle pull requests on GitHub while maintaining traceability with Azure boards <a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>. The integration also enables Copilot code reviews using CodeQL for security scanning, ensuring the generated code is secure without requiring separate CodeQL licensing <a class=\"medius-aisummary\" data-starttime=\"00:11:11\">00:11:11</a>. Next, he elaborates on bridging GitHub Advanced Security with Azure boards by creating “Security Campaigns” <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>, linking campaigns and alerts as Azure work items that can be assigned to Copilot agents for auto fixes <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. This system allows comprehensive tracking of security issues alongside normal development tasks. He further demonstrates how Copilot CLI uses the same MCP settings as the app, enabling developers to query code repositories, identify project tech stacks, and check work item statuses directly from the command line <a class=\"medius-aisummary\" data-starttime=\"00:16:01\">00:16:01</a>.</p>\n\n<p><strong>Infrastructure Updates and Migration Features:</strong> Dan Hellem resumes discussion at <a class=\"medius-aisummary\" data-starttime=\"00:17:37\">00:17:37</a> and introduces new Azure DevOps MCP servers—local and remote—powering agent integration and Copilot Studio connectivity <a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>. He explains collaboration with Entra to remove client restrictions affecting remote operations. Dan then highlights improvements for large enterprises migrating repositories to GitHub using Enterprise Live Migrator (ELM) <a class=\"medius-aisummary\" data-starttime=\"00:22:01\">00:22:01</a>. The tool enables full fidelity repo transfers with minimal downtime, preserving commits, branches, pull requests, and pipeline connections <a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>. A CLI workflow and the upcoming user interface allow organizations to validate, synchronize, and schedule cutovers with ease. For teams that remain on Azure DevOps, Microsoft continues enhancing features to sustain productivity, adding GitHub Copilot AI capabilities directly into Azure repos and pipelines.</p>\n\n<p><strong>AI Features and Cost Management in Azure DevOps:</strong> Dan unveils new Copilot code reviews for Azure Git repos <a class=\"medius-aisummary\" data-starttime=\"00:28:31\">00:28:31</a>, illustrating the workflow for enabling and configuring it at the organization or repository level to manage costs. He presents a demo where dynamic pipelines run shallow clones, send diffs to Copilot, and append suggestion comments automatically to pull requests <a class=\"medius-aisummary\" data-starttime=\"00:30:40\">00:30:40</a>. Subsequent improvements will incorporate Copilot-specific instructions and prior contextual learning. Billing insights show AI usage charged to Azure subscriptions, visible in cost analysis dashboards based on run volume and complexity <a class=\"medius-aisummary\" data-starttime=\"00:34:01\">00:34:01</a>. He also introduces Copilot Autofix inside GitHub Advanced Security <a class=\"medius-aisummary\" data-starttime=\"00:35:35\">00:35:35</a>, which automatically repairs security issues found by CodeQL, generating pull requests instantly upon detection.</p>\n\n<p><strong>Closing, Previews, and Internal Adoption:</strong> As the session approaches conclusion <a class=\"medius-aisummary\" data-starttime=\"00:38:00\">00:38:00</a>, Dan highlights technical previews available for sign‑up—Enterprise Live Migrator, Copilot Code Reviews, and GitHub Advanced Security Autofix—with rolling rollout plans through the product team. He emphasizes Microsoft’s internal adoption of these migration and AI tools, noting that various engineering teams are actively moving repos to GitHub using the same processes described <a class=\"medius-aisummary\" data-starttime=\"00:39:00\">00:39:00</a>. Dave shares repository links containing all demo videos, slides, and documentation for deeper learning <a class=\"medius-aisummary\" data-starttime=\"00:41:03\">00:41:03</a>. The presenters close with the announcement of newly added Apple Mac pay‑per‑minute Azure Pipelines agents <a class=\"medius-aisummary\" data-starttime=\"00:43:50\">00:43:50</a>, and an invitation to join follow‑up discussions at the Ship and Tell pavilion for further questions <a class=\"medius-aisummary\" data-starttime=\"00:44:06\">00:44:06</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"20:30 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/ce405956-4847-4b45-a5e6-e2fa334f2cfe","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ce405956-4847-4b45-a5e6-e2fa334f2cfe","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ce405956-4847-4b45-a5e6-e2fa334f2cfe","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ce405956-4847-4b45-a5e6-e2fa334f2cfe","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ce405956-4847-4b45-a5e6-e2fa334f2cfe","speakerIds":["1768245294609build26-1774649238388001v8Ts","1768245294609build26-1774649204327001sMxG","1768245294609build26-1772218239738001Vovj"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.465Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ce405956-4847-4b45-a5e6-e2fa334f2cfe","ogTitle":"Azure DevOps meets GitHub, the path to AI powered SDLC","ogDescription":"Azure DevOps and GitHub are better together—and the integration keeps getting smarter. In this demo-heavy session, you’ll see how hybrid patterns that connect GitHub with Azure Boards and Azure Pipelines enable Agentic DevOps. See the all of the newest AI-powered capabilities in Azure DevOps. Plus, hear how Microsoft's engineering teams adopted this approach and what they gained.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1772218239738001Vovj","RegistrantKey":"1768245294609build26-1772218239738001Vovj"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774649204327001sMxG","RegistrantKey":"1768245294609build26-1774649204327001sMxG"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774649238388001v8Ts","RegistrantKey":"1768245294609build26-1774649238388001v8Ts"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK202","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Advanced Security","logicalValue":"GitHub Advanced Security"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"a8d1eef8-532b-479e-a9a9-c4a4d032a1ae","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Anush Elangovan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775240864817001hZBb","langLocale":"en-US","sessionId":"1775240864817001hZBb","sessionInstanceId":"1775240864817001hZBb","sessionCode":"BRKSP93","sessionCodeNormalized":"EN-US-BRKSP93","title":"Build AI across client and cloud with AMD ROCm and Microsoft","sortTitle":"build ai across client and cloud with amd rocm and microsoft","description":"As AI moves from experimentation to production, developers need practical ways to build, test, and optimize workloads across client, cloud, and on-prem environments. Learn how AMD ROCm™ provides a common AI software foundation across Radeon™, Ryzen™ AI, and AMD Instinct™ platforms. See how integrations with PyTorch and ONNX/MIGraphX enable portability with fewer code changes while optimizing performance for each target.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The presentation begins with a warm welcome and audio check from the host at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. The speaker outlines the discussion topics, centering on AMD’s broad software ecosystem and the AMD software stack, particularly the ROCm platform and its integration across devices from client systems to the cloud. The introduction sets the tone by promising an exploration of key themes like execution speed in the era of artificial intelligence, open-source development, and AMD’s unified AI software infrastructure.</p>\n\n<p><strong>Speed of Execution and AI Evolution:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:00:59\">00:00:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:27\">00:03:27</a>, the speaker highlights the central thesis: in modern AI-driven transitions, speed is the major competitive advantage. Unlike previous paradigm shifts in mobile and cloud technology that unfolded over years, today’s changes occur in weeks or even days. The ability to quickly adapt to new AI models and technologies determines success. The talk draws parallels between multi-threading in AMD CPUs and the parallelized workflows empowered by AI agents, emphasizing how execution velocity and continuous feedback loops compound innovation across development pipelines.</p>\n\n<p><strong>AMD’s Hardware and Software Ecosystem:</strong> Moving forward at <a class=\"medius-aisummary\" data-starttime=\"00:03:37\">00:03:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:26\">00:05:26</a>, the speaker delves into AMD’s pervasive hardware presence across desktops, edge devices, and data centers, supported by the ROCm (Radeon Open Compute) software stack. This unified platform connects the entire compute spectrum—from personal laptops to cloud GPUs—creating seamless scalability. AMD’s vision of a “pervasive software layer” is explained as a key enabler for cross-environment AI workloads. The ROCm philosophy is rooted in openness, encouraging developer participation in an open-source environment that accelerates progress for the entire ecosystem. Abstractions like Triton and SDK layer optimizations simplify kernel programming and ensure that developers can focus on intent rather than infrastructure details.</p>\n\n<p><strong>Performance and Unified AI Enhancements:</strong> In the section beginning near <a class=\"medius-aisummary\" data-starttime=\"00:09:34\">00:09:34</a>, the speaker discusses AMD’s recent software advancements such as the new ROCm Execution Provider (EP) for ONNX Runtime Gen AI. Benchmarks show threefold faster prefill performance and 40% improvement in decode times on AMD GPUs, highlighting immediate benefits for developers. The discussion notes support for unified memory on Ryzen AI chips and large-scale models, above 100 billion parameters, marking a significant milestone for local and cloud AI enablement. The focus then shifts to the democratization of model deployment through flexible, use-case-driven routing that balances performance, cost, and privacy requirements.</p>\n\n<p><strong>Demo: Hybrid Model Routing in Action:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:05\">00:16:05</a>, a live demonstration by Satya showcases the VLLM Semantic Router integrated with AMD hardware. Using configuration files defining task complexity, domain, and privacy sensitivity, the system intelligently routes inputs between local and cloud-based models. For example, sensitive data prompts trigger local processing for privacy, while complex analytical prompts are sent to cloud models like GPT-4.1 mini. The demonstration emphasizes flexibility, performance, and responsible data handling as the router dynamically manages model routing via the Laminate open-source server, running efficiently on AMD infrastructure. The audience witnesses real-time routing decisions between privacy-conscious local inference and scalable cloud execution.</p>\n\n<p><strong>Conclusion and Vision for the Future:</strong> The talk closes around <a class=\"medius-aisummary\" data-starttime=\"00:17:11\">00:17:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:16\">00:18:16</a> with the key takeaway that modern software development has evolved into an ecosystem where “software is now tokens and time.” The speaker underscores AMD’s mission to accelerate innovation across the full client-to-cloud continuum through ROCm, open-source tools, and agent-based automation. Emphasizing that future systems will be increasingly “agentic,” the closing message reaffirms AMD’s commitment to enabling high-speed, secure, and open AI computation for all developers. The session concludes with an invitation for questions, wrapping up a forward-looking overview of AMD’s strategy in advancing AI and software integration across the global compute landscape.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:30 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:15:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/82519188-f39b-435f-bcc1-4a48673a91cc","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/82519188-f39b-435f-bcc1-4a48673a91cc","captionFileLink":"https://medius.microsoft.com/video/asset/Caption/82519188-f39b-435f-bcc1-4a48673a91cc","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/82519188-f39b-435f-bcc1-4a48673a91cc","speakerIds":["1768245294609build26-1776351028697001ZmDD"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.845Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/82519188-f39b-435f-bcc1-4a48673a91cc","ogTitle":"Build AI across client and cloud with AMD ROCm and Microsoft","ogDescription":"As AI moves from experimentation to production, developers need practical ways to build, test, and optimize workloads across client, cloud, and on-prem environments. Learn how AMD ROCm™ provides a common AI software foundation across Radeon™, Ryzen™ AI, and AMD Instinct™ platforms. See how integrations with PyTorch and ONNX/MIGraphX enable portability with fewer code changes while optimizing performance for each target.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776351028697001ZmDD","RegistrantKey":"1768245294609build26-1776351028697001ZmDD"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"9753d4a5-ab83-4b96-8ce9-b216aea87922","startDateTime":"2026-06-02T22:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kayla Cinnamon, Craig Loewen, Clint Rutkas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018718349001rCjG","langLocale":"en-US","sessionId":"1774018718349001rCjG","sessionInstanceId":"1774018718349001rCjG","sessionCode":"BRK261","sessionCodeNormalized":"EN-US-BRK261","title":"Build and ship faster with a developer-optimized experience on Windows","sortTitle":"build and ship faster with a developer-optimized experience on windows","description":"Learn how Windows delivers a streamlined, end-to-end experience through PCs and OS experiences optimized for developers. See our new experiences in action across WSL, PowerToys and your favorite developer tools so you can code with less toil and stay in your workflows. Walk away with repeatable scenarios that you can easily integrate in your daily development workflows, helping you scale your AI projects with confidence.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Team Overview:</strong> The session opens with hosts Kayla Cinnamon and Craig Loewen welcoming attendees to the \"Build and Ship Faster with Developer-Optimized Experience on Windows\" talk at <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>. Kayla introduces their focus on popular Windows developer tools—Windows Terminal, PowerToys, winget, and WSL. The session aims to showcase how improvements in these tools accelerate both building on and for Windows. The introduction concludes by framing two key themes: enhancing the Windows development environment itself and enabling smoother creation of Windows applications through new and open-source tooling.</p>\n\n<p><strong>Developer Setup and Productivity Enhancements:</strong> Early in the demo portion <a class=\"medius-aisummary\" data-starttime=\"00:01:14\">00:01:14</a>, Kayla demonstrates new features that simplify setting up a Windows development environment. A new winget configuration file automates installing core tools—Ubuntu on WSL, Git, VS Code, Copilot CLI, and other essentials—ensuring idempotent setup so repeated runs only verify presence. She highlights returning customization options like movable taskbars and introduces a redesigned Run dialog built on the PowerToys Command Palette architecture <a class=\"medius-aisummary\" data-starttime=\"00:03:05\">00:03:05</a>. The team also debuts “Intelligent Terminal,” a companion workflow where AI agents such as GitHub Copilot collaborate within the console without overtaking the user’s prompt. This open-source experiment allows developers to select their preferred model and provide feedback directly through GitHub.</p>\n\n<p><strong>Advancing WSL and Core Utilities:</strong> Craig then showcases new WSL container capabilities <a class=\"medius-aisummary\" data-starttime=\"00:06:03\">00:06:03</a>. A new `wsl c` command provides integrated Linux container management, enabling users to pull and run distributions like Ubuntu or Debian without extra tooling. He demonstrates building and running a sample “MarkItDown” container service that converts files to Markdown, using port mapping between Linux and Windows. The team also introduces Windows Coreutils—native ports of 165 standard Linux utilities such as grep and tail—now running seamlessly on Windows <a class=\"medius-aisummary\" data-starttime=\"00:07:23\">00:07:23</a>. Together, these updates strengthen interoperability, offering developers familiar pipelines and Linux tooling while staying in a Windows-native environment. Kayla later highlights the “Comfort Shell,” a curated Ubuntu setup preloaded with favorites like ZSH, Homebrew, and Starship to enhance command-line productivity.</p>\n\n<p><strong>Building for Windows and Developer Tools:</strong> Transitioning to building Windows-specific apps <a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>, Kayla presents the new WinApp CLI—a unified open-source interface for managing Windows SDKs, packaging, signing, and publishing. It integrates with WinGet for easy installation. She adds that developers can extend agent workflows using the Win-Dev-Skills plugin, enabling Copilot or other AI agents to generate, test, and package WinUI 3 apps intelligently. A live demo shows Copilot building a WinUI “MarkItDown” app via voice command using autopilot mode in under a few minutes <a class=\"medius-aisummary\" data-starttime=\"00:17:50\">00:17:50</a>. Additional highlights include PowerToys’ newly added “Grab and Move” utility for easier window management and a strong emphasis on pairing automated build flows with meaningful AI assistance across modern app development.</p>\n\n<p><strong>Performance Optimization with SPGO:</strong> Jianye Lu follows with an in-depth look at Sample-Based Profile Guided Optimization (SPGO) for Windows builds <a class=\"medius-aisummary\" data-starttime=\"00:20:30\">00:20:30</a>. Unlike traditional PGO that requires instrumented binaries and slow profiling runs, SPGO leverages lightweight hardware counters from release builds to capture performance data efficiently. Demonstrating on a small virtual machine project, Jianye walks through using xperf to collect performance traces and converting them into compiler-usable profiles, resulting in a 33% runtime improvement in a Fibonacci computation example <a class=\"medius-aisummary\" data-starttime=\"00:28:20\">00:28:20</a>. Real-world applications like Adobe Photoshop have reported 20% gains using this method. The lesson emphasized is integrating SPGO iteratively in build pipelines for continuous optimization—a scalable, sustainable approach to improving app responsiveness and compile-time efficiency.</p>\n\n<p><strong>Container API Integration and Closing Remarks:</strong> Returning onstage <a class=\"medius-aisummary\" data-starttime=\"00:33:03\">00:33:03</a>, Craig introduces the new WSL container API, allowing developers to embed Linux-backed containers directly inside Windows apps through a NuGet package. Using his MarkItDown service as a showcase, he compiles a WinUI front-end that calls Linux code behind the scenes—running Python services within a WSL-powered virtual machine, complete with port mapping and resource management. Integration with dev containers and third-party tools like VS Code is also confirmed. The session concludes with a story about partnering with DreamWorks’ MoonRay renderer, adapted to Windows via WSL containers <a class=\"medius-aisummary\" data-starttime=\"00:38:00\">00:38:00</a>. Finally, Kayla wraps up by recognizing over 16,000 GitHub contributors across projects like Terminal, PowerToys, and WSL, encouraging attendees to explore the provided links and repos <a class=\"medius-aisummary\" data-starttime=\"00:40:44\">00:40:44</a>. The team thanks the community before moving into Q&A.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"22:45 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/9574001f-c87e-401e-9e24-87990c6318f9","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9574001f-c87e-401e-9e24-87990c6318f9","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9574001f-c87e-401e-9e24-87990c6318f9","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9574001f-c87e-401e-9e24-87990c6318f9","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9574001f-c87e-401e-9e24-87990c6318f9","speakerIds":["1768245294609build26-1769224980177001oYZY","1768245294609build26-1758133761538001NuMh","1768245294609build26-1774284989943001pdAm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.858Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9574001f-c87e-401e-9e24-87990c6318f9","ogTitle":"Build and ship faster with a developer-optimized experience on Windows","ogDescription":"Learn how Windows delivers a streamlined, end-to-end experience through PCs and OS experiences optimized for developers. See our new experiences in action across WSL, PowerToys and your favorite developer tools so you can code with less toil and stay in your workflows. Walk away with repeatable scenarios that you can easily integrate in your daily development workflows, helping you scale your AI projects with confidence.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"C++","logicalValue":"C++"},{"displayValue":"C","logicalValue":"C"}],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758133761538001NuMh","RegistrantKey":"1768245294609build26-1758133761538001NuMh"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774284989943001pdAm","RegistrantKey":"1768245294609build26-1774284989943001pdAm"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769224980177001oYZY","RegistrantKey":"1768245294609build26-1769224980177001oYZY"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK261","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Azure Linux","logicalValue":"Azure Linux"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Windows Development","logicalValue":"Windows Development"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Developer Languages","logicalValue":"Developer Languages"},{"displayValue":"WSL","logicalValue":"WSL"},{"displayValue":"Terminal","logicalValue":"Terminal"}]},{"@search.score":1.0,"documentId":"d1093893-283a-47ce-a391-6abe86e72611","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jordi Janer, Aditi Narvekar, Anastasiya Tarnouskaya, Tucker Burns","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018718047001rVyM","langLocale":"en-US","sessionId":"1774018718047001rVyM","sessionInstanceId":"1774018718047001rVyM","sessionCode":"BRK260","sessionCodeNormalized":"EN-US-BRK260","title":"Build Apps w/ Local AI for Unmetered Intelligence on every Windows PC","sortTitle":"build apps w/ local ai for unmetered intelligence on every windows pc","description":"Start with solution-centric Windows AI APIs - now expanding beyond Copilot+ devices. Use Foundry Local to run open-source models locally. With new tooling in Foundry Toolkit, optimize and prep your models for local AI deployments. Run custom AI workloads locally across GPU, NPU, or CPU with Windows ML, now with support for web apps through WebNN. Learn how the platform enables on-device inference across all Windows PCs to help you ship performant, scalable, and secure AI-powered apps on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Vision for Local AI:</strong> The session begins with Anastasiya Tarnouskaya and Aditi Narvekar welcoming the audience to day two of Build <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>. They introduce the focus of the presentation: building AI-powered apps locally across more than a billion Windows devices without the need for cloud resources. Their discussion stresses a shift toward “unmetered intelligence,” where AI workloads run natively on devices, leveraging the processing power of CPUs, GPUs, and NPUs for performance and energy efficiency <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>. They explain that this approach empowers developers to create scalable, privacy-preserving AI solutions and then preview various demos showing local AI experiences enabled through Foundry on Windows.</p>\n\n<p><strong>Benefits of Local AI and Foundry Overview:</strong> Aditi and Anastasiya outline why developers and enterprises are increasingly turning to local AI—improved privacy, reduced latency, offline functionality, and lower operational costs <a class=\"medius-aisummary\" data-starttime=\"00:03:10\">00:03:10</a>. They introduce Microsoft Foundry on Windows as the unifying environment for building these solutions, encompassing turnkey Windows AI APIs, the Foundry Local platform for running pre-optimized open-source models, and Windows ML for custom deployments <a class=\"medius-aisummary\" data-starttime=\"00:04:19\">00:04:19</a>. To illustrate, they unveil the “Unmetered Token Café,” a hypothetical small business that uses Foundry to enhance operations with AI. This café serves as a running example for subsequent demos that show how AI can be embedded directly into everyday applications without cloud dependencies.</p>\n\n<p><strong>Demonstrations: Foundry APIs in Action:</strong> The first demo showcases a voice-driven ordering system that uses a new speech recognition API to convert spoken input into text, operating locally on a Surface device with a Qualcomm NPU <a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>. Aditi then demonstrates the small language model Phi Silica, now capable of running on the GPU, structuring unstructured data into JSON <a class=\"medius-aisummary\" data-starttime=\"00:08:35\">00:08:35</a>. These examples highlight the versatility of Windows AI APIs and their expanded support for CPU, GPU, and NPU. The session further provides a survey of available task-specific, customization, and media APIs such as text summarization, image description, and video super resolution. Guest presenter Alfred from ClipChamp joins to demonstrate AI-powered video upscaling using local models on an AMD NPU, showing sharper video results and emphasizing low-power, high-quality performance <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>.</p>\n\n<p><strong>Advanced Models and Foundry Local:</strong> The narrative transitions to Microsoft Aion, introduced as the next-generation successor to Phi Silica, available through AI APIs with improved speed and language quality <a class=\"medius-aisummary\" data-starttime=\"00:14:30\">00:14:30</a>. Foundry Local reaches general availability, allowing developers to run open-source models like Qwen 3.5 Vision locally on various hardware. To illustrate, a demo shows an inventory app that uses this model to identify café products from images <a class=\"medius-aisummary\" data-starttime=\"00:17:01\">00:17:01</a>. Foundry Local abstracts away device-specific complexities through its SDK and CLI, offering smooth hardware acceleration across silicon partners. The presenters discuss seamless model management, multi-device compatibility, and demonstrate how developers can experiment via GitHub resources and the Foundry Local command-line interface.</p>\n\n<p><strong>Windows ML, Model Optimization, and Web Integration:</strong> Anastasiya pivots to cover Windows ML as the foundation for running custom models trained or optimized by developers <a class=\"medius-aisummary\" data-starttime=\"00:19:59\">00:19:59</a>. She debuts the new Windows ML CLI <a class=\"medius-aisummary\" data-starttime=\"00:21:32\">00:21:32</a>, which simplifies model conversion, optimization, and benchmarking in one unified tool. A live demo shows optimization steps using a sentiment analysis model, demonstrating how the CLI verifies operator support, optimizes with configuration files, and measures throughput efficiency on an Intel-based Surface Laptop. The discussion then extends to deploying these optimized models with WebNN, enabling near-native AI inference inside browsers using ONNX Runtime Web APIs <a class=\"medius-aisummary\" data-starttime=\"00:30:01\">00:30:01</a>. A demo proves local, low-latency sentiment classification performance across CPUs and NPUs within a web app, representing fully local AI capability even in browser environments.</p>\n\n<p><strong>Partner Demos, Announcements, and Closing Takeaways:</strong> The final portion highlights real-world partner implementations of Windows ML, including Adobe, Canva, and VoiceMod <a class=\"medius-aisummary\" data-starttime=\"00:34:42\">00:34:42</a>. VoiceMod’s team demonstrates live, sub-45ms latency voice transformations powered by Windows ML on Qualcomm NPUs <a class=\"medius-aisummary\" data-starttime=\"00:35:22\">00:35:22</a>, showcasing cross-silicon scalability. New Windows ML features for 2026 are announced, including speed improvements up to 2.6x for generative AI, the availability of Windows ML 2.0 with enhanced ONNX Runtime support, and expanded plugin compatibility from AMD, Intel, Nvidia, and Qualcomm <a class=\"medius-aisummary\" data-starttime=\"00:38:36\">00:38:36</a>. The session concludes by reiterating key lessons from the Unmetered Token Café story: developers can now effortlessly build, optimize, and deploy local AI experiences on any Windows device using Foundry, Windows AI APIs, and Windows ML. Attendees are invited to explore resources, provide feedback, and connect with the speakers after the session <a class=\"medius-aisummary\" data-starttime=\"00:41:50\">00:41:50</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"16:00 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/04c29fa8-c492-43f5-bb90-6582b474053b","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/04c29fa8-c492-43f5-bb90-6582b474053b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/04c29fa8-c492-43f5-bb90-6582b474053b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/04c29fa8-c492-43f5-bb90-6582b474053b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/04c29fa8-c492-43f5-bb90-6582b474053b","speakerIds":["1768245294609build26-1779144962550001mXc0","1768245294609build26-1774284995682001p4go","1768245294609build26-1774386059185001OLB7","1768245294609build26-1774284987167001p2gU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.857Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/04c29fa8-c492-43f5-bb90-6582b474053b","ogTitle":"Build Apps w/ Local AI for Unmetered Intelligence on every Windows PC","ogDescription":"Start with solution-centric Windows AI APIs - now expanding beyond Copilot+ devices. Use Foundry Local to run open-source models locally. With new tooling in Foundry Toolkit, optimize and prep your models for local AI deployments. Run custom AI workloads locally across GPU, NPU, or CPU with Windows ML, now with support for web apps through WebNN. Learn how the platform enables on-device inference across all Windows PCs to help you ship performant, scalable, and secure AI-powered apps on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"C++","logicalValue":"C++"}],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774284987167001p2gU","RegistrantKey":"1768245294609build26-1774284987167001p2gU"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774386059185001OLB7","RegistrantKey":"1768245294609build26-1774386059185001OLB7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995682001p4go","RegistrantKey":"1768245294609build26-1774284995682001p4go"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779144962550001mXc0","RegistrantKey":"1768245294609build26-1779144962550001mXc0"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK260","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"}]},{"@search.score":1.0,"documentId":"06146e2d-24c6-402c-ae97-f3f78ee8e907","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Amanda Silver, Marco Casalaina","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018700487001rXwu","langLocale":"en-US","sessionId":"1774018700487001rXwu","sessionInstanceId":"1774018700487001rXwu","sessionCode":"BRK240","sessionCodeNormalized":"EN-US-BRK240","title":"Build context-aware agents: From data to decisions","sortTitle":"build context-aware agents: from data to decisions","description":"High performance agents are built on intelligence that brings together context, enterprise data, orchestration, and governance. Learn how Foundry IQ, Fabric IQ, and Work IQ provide the enterprise intelligence layer for AI agents. Design agents that can search across organizational knowledge, reason over business data, and operate with awareness of people and work signals. Take action within trusted boundaries, providing a practical foundation for building scalable and reliable agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Industry Shift:</strong>  \nAt the opening of the session <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>, Amanda Silver welcomes attendees and introduces Marco Castelan. Amanda reflects on her two decades at Microsoft, describing a major transformation where software is evolving from human-driven applications to intelligent agents defined by intent and measured by outcomes. These agents can reason, retrieve context, make decisions, and act for users <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>. However, she notes a challenge — over 40% of agentic AI projects fail not because of weak models but due to missing context. Developers often must integrate fragmented data and permissions manually, leading to brittle, unreliable experiences <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>. This sets the stage for the discussion on Microsoft’s new approach to solving this gap.</p>\n\n<p><strong>Introducing Microsoft IQ:</strong>  \nAmanda transitions to explaining Microsoft IQ <a class=\"medius-aisummary\" data-starttime=\"00:02:43\">00:02:43</a> as a context layer for the agentic web. IQ unifies four distinct sources: Work IQ for organizational context, Fabric IQ for business knowledge, Foundry IQ for enterprise knowledge and policy, and Web IQ for global, real-world data. Together, they provide agents a shared understanding of the enterprise and external environments <a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>. When organizations get this right, she says, agents act with precision and reliability—akin to seasoned employees—because they operate with a holistic awareness of people, workflows, and data. Amanda then introduces Marco’s demo to illustrate how IQ works in practice.</p>\n\n<p><strong>Demo: Building and Grounding Agents with IQs:</strong>  \nMarco begins his demonstration <a class=\"medius-aisummary\" data-starttime=\"00:03:35\">00:03:35</a>, starting with Work IQ within Copilot CLI, where an email from Amanda triggers a plan to build a refund agent. The demo encounters initial network setbacks, showing real developer challenges. Despite that, Marco highlights how Work IQ enables authorization and contextual e-mail retrieval. Later, Amanda elaborates on Web IQ <a class=\"medius-aisummary\" data-starttime=\"00:06:24\">00:06:24</a>, which connects agents to real-time verified web information at high speed and efficiency. She cites NASDAQ’s use of Web IQ for real-time executive decision-making as a proof point. Amanda continues with Foundry IQ <a class=\"medius-aisummary\" data-starttime=\"00:10:01\">00:10:01</a>, explaining how it unifies enterprise data sources and enriches information, reducing engineering overhead while maintaining security. Sitecore’s example shows how Foundry IQ powers consistent marketing intelligence across teams.</p>\n\n<p><strong>Fabric IQ and Context Delegation:</strong>  \nThe discussion shifts to Fabric IQ <a class=\"medius-aisummary\" data-starttime=\"00:16:21\">00:16:21</a>, which gives agents real-time understanding of business operations through unified data layers, semantic models, and ontologies. Marco demonstrates how ontologies help translate business entities (like packages or customers) into meaningful structures for agents to reason over. Even through technical difficulties, he illustrates that ontologies define relationships and verbs that agents can interpret semantically, connecting these through Fabric Data Agents for context delegation <a class=\"medius-aisummary\" data-starttime=\"00:26:00\">00:26:00</a>. Amanda adds examples like Q Cells, which use Fabric IQ to manage energy optimization in AI data centers, keeping human oversight integral while empowering autonomous optimization.</p>\n\n<p><strong>Work IQ and Agent Autonomy:</strong>  \nAmanda brings focus back to Work IQ <a class=\"medius-aisummary\" data-starttime=\"00:27:25\">00:27:25</a>, describing it as the people and collaboration layer of Microsoft IQ. It draws knowledge from M365 communication channels—emails, Teams, and documents—so agents understand business context and roles. She explains its core features: chat, text, tools, and workspaces <a class=\"medius-aisummary\" data-starttime=\"00:30:19\">00:30:19</a>. These enable conversation, contextual retrieval, and long-running collaboration among agents and users. Work IQ operates securely within M365, using existing permissions rather than copying data, making agents naturally governance-ready. She highlights Miro’s integration with Work IQ, enabling seamless collaboration without data copies. Marco later resumes his demo, now successful, showing a refund processor agent with its own Teams identity and secure permissions that can read and reply to emails autonomously <a class=\"medius-aisummary\" data-starttime=\"00:35:35\">00:35:35</a>.</p>\n\n<p><strong>Conclusion and Call to Action:</strong>  \nIn closing <a class=\"medius-aisummary\" data-starttime=\"00:39:52\">00:39:52</a>, Amanda summarizes the four IQ layers—Work, Fabric, Foundry, and Web—as the foundation of agentic intelligence and trustworthy automation. She emphasizes that the future lies in providing agents the right context efficiently, not just more data. Microsoft IQ ensures identity, security, and governance are built-in so agent decisions remain explainable and compliant at scale <a class=\"medius-aisummary\" data-starttime=\"00:42:00\">00:42:00</a>. Amanda concludes by inviting developers to start building with Microsoft IQ through labs, GitHub, and an Agents League hackathon <a class=\"medius-aisummary\" data-starttime=\"00:43:18\">00:43:18</a>, encouraging innovators to turn ideas into agentic applications. Marco thanks the audience, ending on a light note about overcoming demo obstacles and celebrating the dawn of a new era in intelligent software.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:00 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","speakerIds":["1768245294609build26-1768930557792001TyJw","1768245294609build26-1758304897855001h9Qq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.455Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/90e4bcf4-03cd-48e6-b0ae-fd4657a5e2f9","ogTitle":"Build context-aware agents: From data to decisions","ogDescription":"High performance agents are built on intelligence that brings together context, enterprise data, orchestration, and governance. Learn how Foundry IQ, Fabric IQ, and Work IQ provide the enterprise intelligence layer for AI agents. Design agents that can search across organizational knowledge, reason over business data, and operate with awareness of people and work signals. Take action within trusted boundaries, providing a practical foundation for building scalable and reliable agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758304897855001h9Qq","RegistrantKey":"1768245294609build26-1758304897855001h9Qq"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1768930557792001TyJw","RegistrantKey":"1768245294609build26-1768930557792001TyJw"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK240","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Advanced Security","logicalValue":"GitHub Advanced Security"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Work IQ","logicalValue":"Work IQ"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"},{"displayValue":"Enterprise","logicalValue":"Enterprise"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"0827c9be-ca7e-4aec-b986-db2c681a5256","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Neta Haiby, Kendra Springer, Ray Zhong","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018716723001rkrH","langLocale":"en-US","sessionId":"1774018716723001rkrH","sessionInstanceId":"1774018716723001rkrH","sessionCode":"BRK251","sessionCodeNormalized":"EN-US-BRK251","title":"Build secure and enterprise-ready agents with Agent 365","sortTitle":"build secure and enterprise-ready agents with agent 365","description":"As AI agents move to production, developers need observability, governance, and security across the agent lifecycle. We'll show how to build enterprise-ready agents using Agent 365 SDK and Purview SDK to instrument runtime visibility, enforce identity-aware access, protect data, and apply policy-based governance. Through practical examples, you'll learn to meet security and compliance requirements without slowing development. Ship agents that are powerful and trustworthy by default.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with Neda welcoming the audience and encouraging everyone to use headphones for clarity <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. She introduces herself and explains that the purpose of the presentation is to show how to build secure, enterprise-ready, and governed agents using Agent 365 <a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>. After an interactive segment asking attendees about their use of AI agents, Neda frames the problem: while many are developing or using agents, few are confident they’re secure or compliant <a class=\"medius-aisummary\" data-starttime=\"00:00:33\">00:00:33</a>. She highlights IDC’s prediction that by 2028 there will be over 1.3 billion agents in organizations, emphasizing the urgency of ensuring observability, governance, and security <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>. Neda concludes her introduction by describing the growing landscape of SaaS agents, endpoint agents, and cloud-based agents and poses key governance questions that lead into the introduction of Agent 365 <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>.</p>\n\n<p><strong>Overview of Agent 365 Capabilities:</strong> Kendra takes over to explain why Agent 365 is essential and what value it provides for both Microsoft and third-party custom agents <a class=\"medius-aisummary\" data-starttime=\"00:02:40\">00:02:40</a>. She introduces its three value pillars: Observe, Govern, and Secure. Observation ensures you can see and analyze all deployed agents across various platforms and understand usage and adoption trends <a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>. Governance helps prevent risk while encouraging innovation by applying guardrails to maintain agent safety without slowing development <a class=\"medius-aisummary\" data-starttime=\"00:04:06\">00:04:06</a>. Security integrates with Defender, Purview, and Entra to provide identity management, threat detection, and compliance monitoring <a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>. She also introduces the Agent 365 SDK, which allows developers to wrap existing agents—Microsoft or third party—for discoverability, visibility, and policy enforcement within Agent 365 <a class=\"medius-aisummary\" data-starttime=\"00:06:10\">00:06:10</a>. The SDK provides agents with an ID and observability while respecting existing governance and security frameworks.</p>\n\n<p><strong>Architecture and Demo Setup:</strong> Kendra elaborates that Agent 365 integrates seamlessly with trusted Microsoft enterprise tools—Entra for identity, Defender for risk protection, Purview for data governance, and Intune for shadow AI detection <a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>. She emphasizes that it manages a heterogeneous mix of agents across different clouds, frameworks, and providers <a class=\"medius-aisummary\" data-starttime=\"00:08:07\">00:08:07</a>. She then defines two critical constructs: Agent Blueprints (reusable instruction sets for agents) and Agent Identities (service principal extensions allowing agents to act consistently with user-like permissions) <a class=\"medius-aisummary\" data-starttime=\"00:12:28\">00:12:28</a>. Before transitioning to the demo, she clarifies how authentication models differ—either acting on behalf of users or operating under their own dedicated identity <a class=\"medius-aisummary\" data-starttime=\"00:13:30\">00:13:30</a>. Aarthy is invited on stage to demonstrate how these concepts come to life in practice through a hands-on walkthrough.</p>\n\n<p><strong>End-to-End Demo: Building and Integrating an Agent:</strong> In the live demo, Aarthy plays multiple roles to simulate the agent development lifecycle <a class=\"medius-aisummary\" data-starttime=\"00:13:40\">00:13:40</a>. She begins as a developer building a travel-planning agent using LangChain and Node.js, showing how to prepare it for Agent 365 by applying skills through the SDK <a class=\"medius-aisummary\" data-starttime=\"00:14:09\">00:14:09</a>. Using the Agent 365 CLI and skills, she demonstrates how observability, work IQ servers, and identity registration are automatically configured <a class=\"medius-aisummary\" data-starttime=\"00:15:07\">00:15:07</a>. Once configured, the agent is deployed, assigned tenant visibility, and activated via the Admin Center. Switching to the user perspective, she interacts with the same agent through Microsoft Teams, requests travel recommendations, and collaborates seamlessly across Teams and Word <a class=\"medius-aisummary\" data-starttime=\"00:19:07\">00:19:07</a>. The demo visually proves how an agent can behave like an enterprise “employee” with secure identities, managed permissions, and transparent telemetry <a class=\"medius-aisummary\" data-starttime=\"00:22:01\">00:22:01</a>.</p>\n\n<p><strong>Admin Experience and Automation:</strong> After Aarthy’s demo, Kendra returns to showcase the administrator view within the Microsoft 365 Admin Center <a class=\"medius-aisummary\" data-starttime=\"00:25:03\">00:25:03</a>. She demonstrates granular observability dashboards showing analytics like total agents, usage stats, and exceptions, as well as the ability to detect orphaned agents and manage ownership assignments <a class=\"medius-aisummary\" data-starttime=\"00:26:26\">00:26:26</a>. She also walks through the Registry Sync feature, which ingests agents from third-party platforms like Google Vertex and Amazon Bedrock for unified visibility <a class=\"medius-aisummary\" data-starttime=\"00:36:31\">00:36:31</a>. Moreover, she introduces Templates—comprehensive, reusable policy bundles combining Entra, Defender, and Purview controls—and demonstrates an agent publishing flow applying those templates automatically during approval <a class=\"medius-aisummary\" data-starttime=\"00:33:49\">00:33:49</a>. She notes new automation Rules that proactively reassign agents or block risky ones without human intervention <a class=\"medius-aisummary\" data-starttime=\"00:33:33\">00:33:33</a>. The visual elements highlight end-to-end control, compliance, and insights across all agents, regardless of origin or platform.</p>\n\n<p><strong>Partner Integration and Conclusion:</strong> The session closes with a partner spotlight featuring Ray from Genspark <a class=\"medius-aisummary\" data-starttime=\"00:39:44\">00:39:44</a>. Ray explains that Genspark integrated its Jasma AI workspace with Agent 365 to unify identity management, security, and observability <a class=\"medius-aisummary\" data-starttime=\"00:41:11\">00:41:11</a>. He describes how this combination allows every agent action to authenticate through Microsoft Entra and log securely to Purview, while leveraging Microsoft tools like Teams, Outlook, and OneDrive through MCP connectors <a class=\"medius-aisummary\" data-starttime=\"00:43:59\">00:43:59</a>. The integration enables flexible back-end compute options while maintaining centralized enterprise governance. Wrapping up, Neda thanks participants and summarizes that Agent 365 empowers organizations to confidently deploy, observe, and secure agents at enterprise scale—whether built internally or externally—making them governed, secure, and enterprise-ready <a class=\"medius-aisummary\" data-starttime=\"00:47:02\">00:47:02</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"18:30 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/1cdd1849-8b45-4ee1-9671-e766fb044081","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/1cdd1849-8b45-4ee1-9671-e766fb044081","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/1cdd1849-8b45-4ee1-9671-e766fb044081","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/1cdd1849-8b45-4ee1-9671-e766fb044081","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/1cdd1849-8b45-4ee1-9671-e766fb044081","speakerIds":["1768245294609build26-1775086106647001Jl6g","1768245294609build26-1770757847067001DhQT","1768245294609build26-1778663825127001Ttju"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.855Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/1cdd1849-8b45-4ee1-9671-e766fb044081","ogTitle":"Build secure and enterprise-ready agents with Agent 365","ogDescription":"As AI agents move to production, developers need observability, governance, and security across the agent lifecycle. We'll show how to build enterprise-ready agents using Agent 365 SDK and Purview SDK to instrument runtime visibility, enforce identity-aware access, protect data, and apply policy-based governance. Through practical examples, you'll learn to meet security and compliance requirements without slowing development. Ship agents that are powerful and trustworthy by default.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775086106647001Jl6g","RegistrantKey":"1768245294609build26-1775086106647001Jl6g"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770757847067001DhQT","RegistrantKey":"1768245294609build26-1770757847067001DhQT"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778663825127001Ttju","RegistrantKey":"1768245294609build26-1778663825127001Ttju"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Purview","logicalValue":"Purview"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Foundry Control Plane","logicalValue":"Foundry Control Plane"},{"displayValue":"Entra","logicalValue":"Entra"}]},{"@search.score":1.0,"documentId":"40707b3c-9771-4908-82e2-507beb5b54cd","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yina Arenas, Naomi Moneypenny, Sharmila Chockalingam","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018712399001rOue","langLocale":"en-US","sessionId":"1774018712399001rOue","sessionInstanceId":"1774018712399001rOue","sessionCode":"BRK230","sessionCodeNormalized":"EN-US-BRK230","title":"Build smarter AI systems in Foundry as models and costs evolve","sortTitle":"build smarter ai systems in foundry as models and costs evolve","description":"Discover how to quickly choose, integrate, and validate AI models inside Microsoft Foundry. Learn techniques for navigating thousands of model options, benchmarking performance, and streamlining your workflow with deep IDE support. Build faster, ship smarter, and stay on top of the evolving AI landscape.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context Setting:</strong> At the start of the session <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>, Yina Arenas opens “Building Smart AI Systems in Foundry” by outlining the challenges of developing AI applications in a constantly evolving landscape. Joined by Naomi Moneypenny, she explains that modern teams are no longer just linking models to chat apps but creating intelligent, agentic systems that improve continuously and require cost awareness, model governance, and scalability. Foundry is introduced as Microsoft’s unified platform for building, deploying, and managing AI systems across their entire lifecycle—with integrated tools such as GitHub Copilot CLI, Fabric IQ, Foundry IQ, and Web IQ. Yina stresses that evaluation and continuous improvement must be embedded early in the process, moving quality assurance to the beginning of the development flow.</p>\n\n<p><strong>Model Ecosystem and Selection in Foundry:</strong> Naomi picks up around <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a> to introduce Foundry’s extensive model catalog and selection strategy. Developers can explore over 11,000 models—ranging from frontier models like OpenAI’s GPT family and Anthropic’s Claude, to open-source offerings via Hugging Face and specialized NVIDIA models—through a consistent API with enterprise-grade curation. New announcements include Claude now running on Azure GB300 hardware and the inclusion of Microsoft’s own multimodal models like Image 2.5. Foundry supports both production-ready architectures and experimental research through Foundry Labs, encouraging flexibility between general-purpose and domain-specific models such as Aurora 1.5 for geospatial intelligence. Yina demonstrates Foundry’s portal, where users can compare models across providers using leaderboards, model benchmarks, and side-by-side evaluations.</p>\n\n<p><strong>Evaluating, Prototyping, and Routing:</strong> The next section <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a> transitions into an applied scenario: designing a travel planning agent through iterative hill climbing. Naomi explains their baseline evaluation using GPT‑4.1 to establish quality, latency, and cost metrics. Once the initial quality scores show gaps, Yina and Naomi demonstrate how Foundry decomposes complex workflows into smaller “jobs to be done.” They introduce model routing—either through Foundry’s built-in router, which automatically selects optimal models from 28 options including GPT‑5.4 and Claude Opus, or through a custom-built intent router using lightweight models like Mistral Small. The improved system significantly reduces running costs and latency while maintaining output quality. Foundry’s tools and GitHub repositories allow developers to replicate this process, emphasizing that teams should prototype, evaluate, and iterate using their own workloads rather than relying on external benchmarks alone.</p>\n\n<p><strong>Evaluation, Custom Metrics, and Rubric-Based Testing:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>, the presenters focus on defining success metrics and using tailored evaluators. Foundry’s evaluation framework includes built‑in measures for quality, risk, and multi-turn behavior as well as customizable evaluators using prompts, code, or rubrics. Yina announces new “rubric-based evaluators” that automatically generate weighted scoring dimensions, such as policy compliance or escalation accuracy, for specific agents like the demo travel planner. Through examples, they show how model runs are traced and analyzed to identify hallucinations and gaps. Naomi demonstrates grounding evaluators in organizational data—in this case, a corporate travel policy—to ensure domain-specific fairness and compliance. Their results show marked improvement but highlight how tuning evaluators exposes areas where human oversight or specialized training remains necessary.</p>\n\n<p><strong>Optimization, Fine-Tuning, and Performance Engineering:</strong> In the extended optimization segment <a class=\"medius-aisummary\" data-starttime=\"00:28:00\">00:28:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:39:00\">00:39:00</a>, Naomi contextualizes cost, quality, and latency trade‑offs as system design issues rather than single-model choices. She announces new capabilities like explicit prompt caching using Azure Context Cache—giving developers owned, reusable context stores with privacy control—and shows architectural levers such as batching, async jobs, distillation, and structured outputs. To enhance domain adherence, Yina demonstrates managed fine‑tuning and distillation using Foundry’s Supervised Fine‑Tuning and Reinforcement Learning options. With under 100 lines of code, they fine‑tune a smaller GPT‑4.1 Nano model against travel policy data, matching quality targets at lower cost. Foundry offers three tuning tiers—Managed Fine‑Tuning, Serverless API with custom rewards, and full‑control RL—plus newly integrated inferencing through Fireworks (GA) and Managed Compute partnerships with Hugging Face and NVIDIA, ensuring consistent APIs and governance across compute backends.</p>\n\n<p><strong>Operations, Observability, and Continuous Improvement:</strong> In the closing discussion <a class=\"medius-aisummary\" data-starttime=\"00:40:00\">00:40:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:00\">00:45:00</a>, Naomi and Yina transition from optimization to production-grade AI operations. Foundry provides observability through tracing, evaluation, and Azure Monitor integration, enabling cost, latency, and agent behavior monitoring. Developers can query Copilot within Foundry to obtain improvement suggestions and visualize metrics like time to last byte. The summary reiterates that building intelligent, resilient systems demands a holistic approach—select the right models, decompose tasks, evaluate continuously, optimize across quality and cost, and embed governance into CI/CD pipelines. They conclude by directing developers to explore further sessions on agent optimization, safety, and responsible AI, and to start building on Foundry via ai.azure.com. The session ends with applause, underscoring the message that Foundry delivers an evolving, governed AI platform for continuous innovation. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"19:30 - 20:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","speakerIds":["1768245294609build26-1768532909165001FoJl","1768245294609build26-1775680871968001CAxd","1768245294609build26-1752517267116001lWQS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.852Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/fb5d9581-c0ac-4290-9a9b-1fdcb98d6443","ogTitle":"Build smarter AI systems in Foundry as models and costs evolve","ogDescription":"Discover how to quickly choose, integrate, and validate AI models inside Microsoft Foundry. Learn techniques for navigating thousands of model options, benchmarking performance, and streamlining your workflow with deep IDE support. Build faster, ship smarter, and stay on top of the evolving AI landscape.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1768532909165001FoJl","RegistrantKey":"1768245294609build26-1768532909165001FoJl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775680871968001CAxd","RegistrantKey":"1768245294609build26-1775680871968001CAxd"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1752517267116001lWQS","RegistrantKey":"1768245294609build26-1752517267116001lWQS"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK230","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"dba22f16-fccb-4d4b-939d-5078d33a8124","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Peter Steinberger","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701283001rsNq","langLocale":"en-US","sessionId":"1774018701283001rsNq","sessionInstanceId":"1774018701283001rsNq","sessionCode":"BRK245","sessionCodeNormalized":"EN-US-BRK245","title":"Build the thing that builds the thing","sortTitle":"build the thing that builds the thing","description":"The way software is built is changing rapidly. Learn how we build a whole ecosystem of tools to build OpenClaw faster and with more confidence.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Core Philosophy:</strong> The talk begins with the speaker greeting the audience and explaining that this will not be a conventional presentation (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>). The central theme is \"build a thing that builds the thing\" — a concept that emphasizes developing tools and automation that enhance the software-building process. The speaker notes that in the age of coding agents, many traditional ways of writing software have become outdated (<a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>). Instead of focusing solely on speed, developers should now ask how they can help their AI agents build faster and close feedback loops more efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>).</p>\n\n<p><strong>From Open Source Chaos to Automation:</strong> Discussing open source development, the speaker humorously reflects on media reports of “failed” projects, clarifying that each was built to improve their own workflow rather than for public success (<a class=\"medius-aisummary\" data-starttime=\"00:01:28\">00:01:28</a>). When managing their Opencloud project, the number of GitHub issues exploded beyond 10,000 (<a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>), prompting the creation of tools like Close Reaper to automatically review, group, and close issues (<a class=\"medius-aisummary\" data-starttime=\"00:05:23\">00:05:23</a>). By integrating project vision documents and clear automated rules, they reduced clutter and closed around 15,000 issues autonomously (<a class=\"medius-aisummary\" data-starttime=\"00:06:35\">00:06:35</a>). The system even updates weekly through continuous code-agent reevaluation, ensuring lingering issues are resolved without direct human review.</p>\n\n<p><strong>Scaling Tools and Managing Data:</strong> As more maintainers joined the project, the speaker needed better management dashboards. They created a report system integrating both GitHub and Discord discussions (<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>) to highlight true maintenance contribution beyond code commits. To collect data from community chats, they built “Disk Crawl,” a crawler for Discord, and stored structured outputs directly in GitHub repositories for transparency and safe backup (<a class=\"medius-aisummary\" data-starttime=\"00:15:14\">00:15:14</a>). This automation ecosystem extended into tools like Release Bar, which visually tracks version recency to guide release timing (<a class=\"medius-aisummary\" data-starttime=\"00:15:35\">00:15:35</a>), demonstrating how each small helper tool compounds to accelerate overall productivity.</p>\n\n<p><strong>Solving Bottlenecks and Expanding Infrastructure:</strong> A persistent issue was hitting GitHub’s API rate limits when running multiple agents. This led to building “Octopus,” a token-balancing layer using both user and GitHub App tokens to prevent throttling (<a class=\"medius-aisummary\" data-starttime=\"00:17:05\">00:17:05</a>). When tests became computationally expensive, the speaker introduced “Crab Box,” which spins up isolated cloud instances to distribute testing workloads across platforms and providers like AWS or Azure (<a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>). They then added GUI capabilities using VNC and even computer vision support so agents could autonomously interact with user interfaces. Later expansions, such as “Mantis,” let agents record videos of test failures and fixes (<a class=\"medius-aisummary\" data-starttime=\"00:25:41\">00:25:41</a>), turning previously manual verification into automated visualization.</p>\n\n<p><strong>Creating Continuous Review and Collaboration Systems:</strong> To address inefficiencies with manual reviews, “Auto Review” was implemented, allowing coding agents to iteratively self-review until code passes test criteria (<a class=\"medius-aisummary\" data-starttime=\"00:27:12\">00:27:12</a>). This concept expanded into “Core Patch,” wherein large projects are subdivided for parallel agent-based audits (<a class=\"medius-aisummary\" data-starttime=\"00:35:00\">00:35:00</a>). Team collaboration improved with “Crab Fleet,” enabling co-working sessions between maintainers and their agents — effectively turning debugging and coding into multiplayer experiences (<a class=\"medius-aisummary\" data-starttime=\"00:36:35\">00:36:35</a>). Each automation layer derived from specific frustrations, reinforcing a culture where annoyance signals opportunity for optimization and innovation.</p>\n\n<p><strong>Conclusion and Q&A Insights:</strong> The talk closes by encouraging developers to harness irritation as a creative signal to automate recurring pain points (<a class=\"medius-aisummary\" data-starttime=\"00:39:02\">00:39:02</a>). Questions from the audience explore tool dependency, compounding automation benefits, and balancing workload against innovation (<a class=\"medius-aisummary\" data-starttime=\"00:40:02\">00:40:02</a>). The speaker emphasizes that none of these solutions must be polished — “slop is fine” if it works. Small tools and prototypes accelerate future builds and free up focus for higher-level creativity. The entire session reinforces the principle that in the agent-driven era, productivity scales not by writing code faster, but by constantly building the tools that empower both humans and agents to create better systems together.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:30 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:15:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/5a21e754-8418-4276-9e66-5041ed2f019d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/5a21e754-8418-4276-9e66-5041ed2f019d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/5a21e754-8418-4276-9e66-5041ed2f019d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/5a21e754-8418-4276-9e66-5041ed2f019d","speakerIds":["1768245294609build26-1772238022106001lZo4"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.456Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/5a21e754-8418-4276-9e66-5041ed2f019d","ogTitle":"Build the thing that builds the thing","ogDescription":"The way software is built is changing rapidly. Learn how we build a whole ecosystem of tools to build OpenClaw faster and with more confidence.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772238022106001lZo4","RegistrantKey":"1768245294609build26-1772238022106001lZo4"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Purview","logicalValue":"Microsoft Purview"},{"displayValue":"ACA","logicalValue":"ACA"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"}]},{"@search.score":1.0,"documentId":"af53d259-81e3-4431-9376-e13482adca07","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kirupa Chinnathambi, Patrick Nikoletich, Stuart Schaefer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018718648001rUpo","langLocale":"en-US","sessionId":"1774018718648001rUpo","sessionInstanceId":"1774018718648001rUpo","sessionCode":"BRK262","sessionCodeNormalized":"EN-US-BRK262","title":"Building Agents You Can Trust on Windows","sortTitle":"building agents you can trust on windows","description":"Agents now run real commands, modify files, and move data. This breakout highlights the full set of Windows primitives that help agents discover, reason about, and execute AI‑powered activities across the system, while still operating within clear boundaries. Through live demos, you’ll see how agents identify available capabilities, plan and take meaningful actions, and how Windows layers permission scoping, inspection, developer tool capabilities, and rollback to keep developers in control.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with a welcome from Krupa at <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>, setting the stage for a discussion about the evolution of AI agents. She explains that modern agents now act autonomously — calling tools, running workflows, and making decisions without direct human prompts <a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>. With this new level of autonomy comes the challenge of managing safety, predictability, and accountability. She introduces the key focus of the talk: how Windows is building capabilities to ensure AI agents operate securely and reliably. Joining her are Stuart from the Windows Platform team and Patrick from GitHub, who will later demonstrate practical integrations with GitHub Copilot <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>.</p>\n\n<p><strong>Risks of Autonomous Agents:</strong> Krupa continues by addressing the problems emerging as agents gain more power and independence <a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>. She cites real-world failures where agents misinterpreted commands — such as deleting important emails or production databases — illustrating how small errors get amplified into serious consequences <a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>. The discussion highlights that traditional approaches like prompt engineering cannot fully mitigate these risks because agents interact across multiple vectors: with humans, tools, apps, and even other agents <a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:50\">00:02:50</a>. Krupa turns the floor over to Patrick from GitHub, who shares how the Copilot team observed the compounding risks when users remove protective gates and let agents run freely for long sessions in both CLI and SDK environments <a class=\"medius-aisummary\" data-starttime=\"00:03:07\">00:03:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:21\">00:04:21</a>.</p>\n\n<p><strong>Windows Platform Solutions – Identity, Containment, and Manageability:</strong> After defining the risks, Krupa and Stuart explain how Windows now treats agents as a new type of user at the OS level <a class=\"medius-aisummary\" data-starttime=\"00:04:41\">00:04:41</a>. Three essential design principles are introduced: identity, containment, and manageability. Each agent must have a distinct identity separate from the human user <a class=\"medius-aisummary\" data-starttime=\"00:05:20\">00:05:20</a>. It must operate within controlled containment boundaries that limit access and reduce blast radius <a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:44\">00:05:44</a>. Finally, manageability allows IT administrators to set and enforce policies governing what agents can and cannot do. Stuart demonstrates how Windows enforces identity-level separation using a demo showing an Open Clog agent running securely in its own user session, isolated from human desktop processes <a class=\"medius-aisummary\" data-starttime=\"00:08:02\">00:08:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>.</p>\n\n<p><strong>Containment with Microsoft Execution Containers (MXC):</strong> Stuart dives deeper into containment mechanisms with the MXC Library <a class=\"medius-aisummary\" data-starttime=\"00:11:12\">00:11:12</a>. He explains that due to agent non-determinism, the system must dynamically create isolation environments that scale based on risk. MXC provides declarative containerization policies, enabling developers to define how agents operate securely by declaring permissions, file access, network limits, and runtime constraints <a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:14\">00:14:14</a>. Live demos illustrate MXC enforcement — first showing failure due to read-only policies and then success once proper permissions are granted <a class=\"medius-aisummary\" data-starttime=\"00:15:15\">00:15:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:50\">00:15:50</a>. Stuart also presents sandbox monitoring of a malicious agent unsuccessfully trying to break containment, proving how the environment protects critical system resources <a class=\"medius-aisummary\" data-starttime=\"00:16:15\">00:16:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:07\">00:17:07</a>. He concludes with Hyperlite, a hardware-backed micro VM demo that securely executes tasks in milliseconds with minimal footprint <a class=\"medius-aisummary\" data-starttime=\"00:17:12\">00:17:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>.</p>\n\n<p><strong>Manageability, Supervision, and Integration:</strong> The discussion transitions to managing autonomous agents and ensuring oversight <a class=\"medius-aisummary\" data-starttime=\"00:18:21\">00:18:21</a>. Stuart emphasizes policy enforcement, user consent, and supervision so that agents maintain human-in-the-loop transparency. He advises developers to build systems that surface agent intentions and allow real-time user intervention when errors occur <a class=\"medius-aisummary\" data-starttime=\"00:19:03\">00:19:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:56\">00:19:56</a>. Patrick then demonstrates GitHub Copilot’s experimental sandbox integration, showing real-world containment against destructive agent actions like file deletion and unsafe repository exposure <a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:04\">00:24:04</a>. He reveals how users can toggle protection, customize network filters, and observe sandbox restrictions while maintaining developer flexibility across CLI, SDK, and integrated apps. The sandbox improves safety in autonomous workflows while providing developers guided control <a class=\"medius-aisummary\" data-starttime=\"00:25:25\">00:25:25</a>.</p>\n\n<p><strong>Conclusion and Future Outlook:</strong> The session closes with Krupa summarizing the fundamental steps for responsible agent deployment on Windows <a class=\"medius-aisummary\" data-starttime=\"00:26:12\">00:26:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:01\">00:27:01</a>. Each agent should have a unique identity, run inside sandboxed containment, and remain managed through IT policies. She introduces the upcoming integration with Agent 365 built over MXC, which supports enterprise-grade protections through Defender, Entra, and Intune <a class=\"medius-aisummary\" data-starttime=\"00:27:27\">00:27:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:50\">00:27:50</a>. Krupa emphasizes that Windows now serves as a platform not just for running agents, but for ensuring they run safely, predictably, and at scale <a class=\"medius-aisummary\" data-starttime=\"00:28:03\">00:28:03</a>. The talk concludes with an invitation to explore MXC and engage with the Windows, GitHub, and OpenAI teams for deeper collaboration on advancing secure agentic ecosystems <a class=\"medius-aisummary\" data-starttime=\"00:28:19\">00:28:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:31\">00:28:31</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:00 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/ceeb5203-d352-4d45-8be6-9db57724f003","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ceeb5203-d352-4d45-8be6-9db57724f003","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ceeb5203-d352-4d45-8be6-9db57724f003","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ceeb5203-d352-4d45-8be6-9db57724f003","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ceeb5203-d352-4d45-8be6-9db57724f003","speakerIds":["1768245294609build26-1774284992189001pC5a","1768245294609build26-1777003136185001uvnL","1768245294609build26-1776972565872001qoXF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.858Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ceeb5203-d352-4d45-8be6-9db57724f003","ogTitle":"Building Agents You Can Trust on Windows","ogDescription":"Agents now run real commands, modify files, and move data. This breakout highlights the full set of Windows primitives that help agents discover, reason about, and execute AI‑powered activities across the system, while still operating within clear boundaries. Through live demos, you’ll see how agents identify available capabilities, plan and take meaningful actions, and how Windows layers permission scoping, inspection, developer tool capabilities, and rollback to keep developers in control.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284992189001pC5a","RegistrantKey":"1768245294609build26-1774284992189001pC5a"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776972565872001qoXF","RegistrantKey":"1768245294609build26-1776972565872001qoXF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777003136185001uvnL","RegistrantKey":"1768245294609build26-1777003136185001uvnL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"}]},{"@search.score":1.0,"documentId":"be65609a-8cb9-4be6-8747-dcf8ea1b037c","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Glenn Condron, Amanda Foster, Shawn Henry","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018700963001rb5P","langLocale":"en-US","sessionId":"1774018700963001rb5P","sessionInstanceId":"1774018700963001rb5P","sessionCode":"BRK243","sessionCodeNormalized":"EN-US-BRK243","title":"Claw and agent harness in Microsoft Foundry","sortTitle":"claw and agent harness in microsoft foundry","description":"Go deep on multi-agent systems built on Microsoft Foundry, featuring Claw agent patterns and the hosted agents architecture. Explore long-running agents with triggers, state management, and file access—all natively supported on Foundry. See how coding agents built with GitHub Copilot SDK and Claude Agent SDK integrate into multi-agent workflows using Microsoft Agent Framework. Learn how to coordinate, host, and operate these systems with observability and continuous evals.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Sean Henry opens Day 2 of Build, joined by Glenn and Amanda from Microsoft Foundry, the company’s AI platform for building intelligent applications and agents. He outlines that this session will dive deeper than Tina and Jeff’s earlier overview, focusing on advanced agent deployment concepts such as claws and agent harnesses. Sean explains Foundry’s layered architecture—comprising the intelligence layer with available models, the runtime layer for hosting and managing agents, and the human-agent collaboration layer enabling integration with tools like Teams and M365 Copilot. He emphasizes that all of this is supported by strong trust, security, and observability fundamentals within Foundry. Sean concludes this part by previewing that the team will focus on the runtime and collaboration aspects and later explore deployment of more advanced agents beyond simple prompt-based ones.</p>\n\n<p><strong>Defining Agent Harnesses and Core Architecture:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>, Sean introduces the concept of “agent harnesses,” describing them as the outer shell or platform layer surrounding an agent, allowing it to handle long-running and complex tasks. He likens the harness to a car that houses the motor—the core intelligence or “agent loop.” This loop continues until the agent achieves its goal by iteratively processing context, acting through tools, and refining responses. Foundry’s harness manages context growth, compaction, and enrichment through prompts, skills, and messages. Sean points out that agents can perform human-like tasks when given access to computer tools such as file systems and code execution environments. The harness also enables orchestration of specialized agents, memory management across sessions, lifecycle hooks for integration, and human-in-the-loop capabilities for oversight of risky operations. This section builds the foundation for understanding robust, extensible agent design before transitioning into demonstrations of harness deployment.</p>\n\n<p><strong>Hermes Demonstration and Foundry Integration:</strong> Glenn takes the stage at <a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a> to demonstrate deploying the Hermes agent, a claw-like architecture built around autonomous, persistent agents. He explains how Hermes integrates memory, tools, and autonomy, running within sandboxed environments or VMs while maintaining constant communication through platforms like Slack or Telegram. Glenn showcases Hermes running on his terminal, backed by Foundry-hosted agents using Azure credentials and Foundry models. He walks through connecting to SharePoint via MCP tools and highlights the concept of “routines” (<a class=\"medius-aisummary\" data-starttime=\"00:13:55\">00:13:55</a>), automated maintenance tasks that manage skills and backups, allowing agents to self-sustain even when not running continuously. Glenn emphasizes the importance of agent state and sandbox persistence, illustrating how each Hermes instance creates unique environments that improve with use but require thoughtful state recovery strategies. Through this live demo, he shows the balance between autonomy, cost optimization, and resilience when deploying claws-style agents at scale within Foundry.</p>\n\n<p><strong>Building Custom Agents with Microsoft Agent Framework:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:21:21\">00:21:21</a>, the next segment demonstrates how to create custom harnesses using Microsoft Agent Framework, which supports Python and C#. Sean elaborates on its components: the agent loop, workflows, and harnesses. The agent loop allows integration with models and tools from Foundry as well as third-party providers like OpenAI, Anthropic, or Gemini. Workflows enable multi-agent systems that can sequence duties or use supervisory planning, and harnesses add common tools for file access, code execution, context management, and planning. Sean presents a coding demo (<a class=\"medius-aisummary\" data-starttime=\"00:26:05\">00:26:05</a>) showing how easily developers can wrap a harness around an agent, customize capabilities, and visualize operations through a console UI. He further illustrates deployment options—running agents in GUIs via AGUI integration, connecting them to Copilot Kit interfaces for interactive data visualizations, and deploying directly to Foundry where monitoring and evaluation tools track performance. This section underscores Foundry’s flexibility and developer control for building deeply customized AI workflows.</p>\n\n<p><strong>Human Collaboration and Autopilot Agents:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:35:01\">00:35:01</a>, Amanda transitions to how agents appear and operate in collaborative environments. She demonstrates Foundry’s simplified publishing of agents directly into Teams or M365 Copilot, explaining that creators can use a one-click process to configure and share agents enterprise-wide. Amanda then introduces a new generation called “Autopilot agents” (<a class=\"medius-aisummary\" data-starttime=\"00:39:46\">00:39:46</a>), which act autonomously with their own user identities, allowing them to send messages, manage documents, and perform independent actions traditionally limited to human accounts. She showcases a “work stream manager” autopilot sample with onboarding, access controls, and group chat behaviors—agents can respond intelligently to permitted users and manage ongoing discussions. This represents Foundry’s evolution toward persistent, collaborative AI agents seamlessly integrated into everyday productivity tools, bridging individual assistance and organizational automation.</p>\n\n<p><strong>Conclusion and Call to Action:</strong> Sean returns at <a class=\"medius-aisummary\" data-starttime=\"00:44:03\">00:44:03</a> to wrap up the session. He summarizes the three core demonstrations: deploying Hermes as an off-the-shelf harness, building custom harnesses with Microsoft Agent Framework, and publishing agents through Foundry for enterprise collaboration. The team directs attendees to GitHub for code samples and invites them to join the ongoing Agent League Hackathon (<a class=\"medius-aisummary\" data-starttime=\"00:45:00\">00:45:00</a>), highlighting the opportunity to apply these techniques creatively. Sean encourages developers to visit the Foundry booth, connect with the speakers, and share ideas about future agent-building projects. As the presentation closes, the group emphasizes Foundry’s expanding toolkit for developers—from harness design and runtime management to human-AI collaboration—reflecting Microsoft’s vision for scalable, secure, and intelligent agent applications in production environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"18:30 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/ce857844-2448-482c-a66b-efa81e214b35","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ce857844-2448-482c-a66b-efa81e214b35","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ce857844-2448-482c-a66b-efa81e214b35","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ce857844-2448-482c-a66b-efa81e214b35","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ce857844-2448-482c-a66b-efa81e214b35","speakerIds":["1768245294609build26-1775581852702001WrWI","1768245294609build26-1768498845637001TuQh","1768245294609build26-1775078836270001Iwf7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.455Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ce857844-2448-482c-a66b-efa81e214b35","ogTitle":"Claw and agent harness in Microsoft Foundry","ogDescription":"Go deep on multi-agent systems built on Microsoft Foundry, featuring Claw agent patterns and the hosted agents architecture. Explore long-running agents with triggers, state management, and file access—all natively supported on Foundry. See how coding agents built with GitHub Copilot SDK and Claude Agent SDK integrate into multi-agent workflows using Microsoft Agent Framework. Learn how to coordinate, host, and operate these systems with observability and continuous evals.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775078836270001Iwf7","RegistrantKey":"1768245294609build26-1775078836270001Iwf7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768498845637001TuQh","RegistrantKey":"1768245294609build26-1768498845637001TuQh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775581852702001WrWI","RegistrantKey":"1768245294609build26-1775581852702001WrWI"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK243","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Claws","logicalValue":"Claws"},{"displayValue":"Openclaw","logicalValue":"Openclaw"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"e002e239-282c-4bd6-a99b-574270e910fe","startDateTime":"2026-06-03T20:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sachin Patney, Ben Zulauf","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018704951001rHLF","langLocale":"en-US","sessionId":"1774018704951001rHLF","sessionInstanceId":"1774018704951001rHLF","sessionCode":"BRK225","sessionCodeNormalized":"EN-US-BRK225","title":"Data, apps, and agents: the future of app dev with Rayfin","sortTitle":"data, apps, and agents: the future of app dev with rayfin","description":"Apps are coming to the data platform. With Rayfin, your entire backend is expressed in code—and Rayfin translates it into the services it needs: databases, functions, and more. Write the code yourself, or describe your app and let agents generate it for you. Either way, deploy with the CLI as a secure, scalable solution. With native access to Fabric data, analytics, and AI, Rayfin eliminates backend stitching so teams move from idea to impact faster.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction to Agentic Coding and Rayfin:</strong> At the start of the presentation (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>), Ben Zulauf introduces himself and Sachin, sharing their excitement about discussing the future of app development with Rayfin. Ben explains how agentic coding is democratizing software creation, transforming anyone—including non-developers—into app makers. He illustrates this with a personal story about his friend Brent, who used agentic coding to create apps for his kids’ soccer team. This section emphasizes how the technology empowers new creators and enhances productivity for professional developers. However, Ben quickly pivots to note that when moving to enterprise environments with real data and compliance requirements (<a class=\"medius-aisummary\" data-starttime=\"00:01:53\">00:01:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:22\">00:02:22</a>), the informal “vibe coding” approach fails to meet security and governance standards—leading to the need for a more robust framework like Rayfin. </p>\n\n<p><strong>Rayfin Overview and Demo:</strong> Ben introduces Rayfin as the backend solution for the AI coding era (<a class=\"medius-aisummary\" data-starttime=\"00:02:37\">00:02:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>). He explains that Rayfin allows developers and AI agents to define databases, functions, storage, and access policies in code through its SDK and CLI, all deployable to Microsoft Fabric with a single command. The demonstration that follows (<a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:10\">00:08:10</a>) centers around building a delivery application for a fictional company, Zava. Ben showcases how quickly an app can be created and deployed, complete with frontend and backend integration. The resulting app allows delivery drivers to log items and customers to confirm receipt, with data stored securely in Fabric’s SQL database. The demo highlights Rayfin’s capability to streamline development while maintaining enterprise-grade compliance and data structure integrity. </p>\n\n<p><strong>Technical Deep Dive by Sachin:</strong> Sachin Patney then dives into how Rayfin technically works under the hood (<a class=\"medius-aisummary\" data-starttime=\"00:08:10\">00:08:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:02\">00:14:02</a>). He explains entity and property decorators within TypeScript that define database tables, types, constraints, and relationships—all directly mapped to Fabric’s SQL database. Sachin also describes how role-based policies and security can be written directly into code, allowing authenticated users controlled access to their own data. Rayfin automates schema migration when models change, reducing overhead. He outlines the YAML configuration model that supports Microsoft SQL, with future support for PostgreSQL. The flexibility extends to hosting options—frontends and backends can deploy independently, share a common backend, or connect through scopes with varying permissions. Additionally, Sachin details integration through connectors to existing enterprise data in Fabric’s OneLake, enabling read/write operations and type-safe access. </p>\n\n<p><strong>Templates and Data App Innovations:</strong> The presentation then transitions to Rayfin’s templating system (<a class=\"medius-aisummary\" data-starttime=\"00:14:02\">00:14:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:01\">00:18:01</a>), which helps organizations standardize and accelerate app development. Sachin explains how templates can encapsulate frameworks, APIs, and libraries, allowing internal teams to quickly replicate standardized dashboards or tools. Sujata joins to present the “Data App” template co-developed with Microsoft’s Power BI team (<a class=\"medius-aisummary\" data-starttime=\"00:18:01\">00:18:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:41\">00:21:41</a>), showcasing how analytics can now live within application environments. This template combines a data package for simplifying queries, a visual package using React-based enterprise visuals, and agent skills that help coding agents create elegant, purpose-built dashboards. Sujata demonstrates how an analytics app can be generated using minimal prompts from a semantic model, producing enterprise-quality dashboards that combine data insights with actionable controls. </p>\n\n<p><strong>Deployment, Fabric Integration, and Open Source Plans:</strong> Sachin returns to explain why Fabric is an ideal environment for Rayfin apps (<a class=\"medius-aisummary\" data-starttime=\"00:21:52\">00:21:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>). Fabric provides enterprise-grade security and compliance management, storing all app data within OneLake, Microsoft’s unified data lake. This framework allows developers and business users to seamlessly access and analyze application data. Moreover, Fabric’s SaaS model simplifies deployment, letting apps share capacity across workloads. Sachin covers development workflows that support local or cloud-based workspaces and containerized testing. To encourage flexibility, Microsoft will open source parts of the Rayfin runtime, enabling self-hosted deployment on different infrastructures. He highlights how these steps preserve governance while extending customization possibilities. </p>\n\n<p><strong>Partnerships and Future Outlook:</strong> To illustrate Rayfin’s ecosystem growth, Sachin introduces Carl from Replit (<a class=\"medius-aisummary\" data-starttime=\"00:25:55\">00:25:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:11\">00:28:11</a>). Carl demonstrates a Replit integration that empowers non-technical users to build and deploy governed Rayfin applications directly from the browser. Enterprise privacy and Fabric-based data governance are built in. He cites positive user feedback from Leatherman, emphasizing how combining Replit’s rapid development with Fabric’s data compliance fulfills enterprise needs. Ben concludes the session (<a class=\"medius-aisummary\" data-starttime=\"00:28:33\">00:28:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:31:36\">00:31:36</a>) by outlining upcoming releases—support for functions, RBAC, and additional connectors—as well as expanding to B2B and B2C scenarios and offering a free-tier onboarding experience. He invites attendees to experiment with Rayfin via GitHub, sign up for the Replit preview, and contribute feedback as the platform evolves. The talk ends with a call for developers and agents alike to help shape the future of enterprise app development with Rayfin. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"20:30 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","speakerIds":["1768245294609build26-1774284996963001pGpM","1768245294609build26-1774284988298001pfqa"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c1caffa3-94c6-4ba4-92e3-1aef6b91878d","ogTitle":"Data, apps, and agents: the future of app dev with Rayfin","ogDescription":"Apps are coming to the data platform. With Rayfin, your entire backend is expressed in code—and Rayfin translates it into the services it needs: databases, functions, and more. Write the code yourself, or describe your app and let agents generate it for you. Either way, deploy with the CLI as a secure, scalable solution. With native access to Fabric data, analytics, and AI, Rayfin eliminates backend stitching so teams move from idea to impact faster.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284988298001pfqa","RegistrantKey":"1768245294609build26-1774284988298001pfqa"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996963001pGpM","RegistrantKey":"1768245294609build26-1774284996963001pGpM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"a2e8664e-e79b-4c6d-ba67-decefb9dba17","startDateTime":"2026-06-02T22:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alicia Frame, Omkar More, Long Chen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018712707001rIUj","langLocale":"en-US","sessionId":"1774018712707001rIUj","sessionInstanceId":"1774018712707001rIUj","sessionCode":"BRK231","sessionCodeNormalized":"EN-US-BRK231","title":"Deploy. Observe. Learn. Reinforcement learning for production agents","sortTitle":"deploy. observe. learn. reinforcement learning for production agents","description":"Agents don't fail in demos — they fail in production. In this breakout, see how teams use fine-tuning and reinforcement learning on Microsoft Foundry to improve production agents using real usage signals. We cover when fine-tuning reduces cost and latency, when RL delivers deeper gains, and how Foundry makes it easy to train, evaluate, and redeploy safely. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> At the start of the session (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:46\">00:00:46</a>), Alicia Frame welcomes the audience alongside Omkar More and introduces the focus of the talk: fine-tuning and reinforcement learning for production agents. Alicia, the Product Lead for Model Customization, outlines how the team will discuss deploying, observing, and improving AI agents to achieve better-performing models that are cheaper and faster. The agenda includes a short overview of fine-tuning concepts, demonstrations on distillation, supervised fine-tuning, and reinforcement learning, followed by a deep-dive preview of Microsoft’s low-level training API referred to as “PyTorch as a service.” She explains that each demo will show increasingly advanced customization methods culminating in a coding agent example using natural language to define fine-tuning goals.</p>\n\n<p><strong>Foundation: Fine-Tuning Motivation and Context:</strong> Alicia situates the work within Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:02:06\">00:02:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>), explaining that Foundry hosts over 11,000 models forming the base for agent runtimes. She emphasizes the need for evaluation and optimization tools to make agents smarter and more efficient. Fine-tuning emerges as an answer to skyrocketing token consumption and cost in the expanding use of AI agents. Rather than building massive frontier models, organizations increasingly fine-tune smaller, cheaper models that retain comparable quality at reduced cost and latency. This section highlights the transition from improving model quality through domain-specific data to optimizing agents for practical performance, and how smaller models like GPT-4.1 or Qwen versions offer 10–30x cost reductions while providing faster responses in production scenarios.</p>\n\n<p><strong>Distillation Demo and Evaluation:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>, Omkar introduces the first demo: distillation as a route to cheaper and faster models. Using a customer support agent scenario for retail refunds, he demonstrates how Foundry captures agent traces while processing customer requests, including tool usage such as fetching order details or fulfillment status. These traces become datasets for training smaller models. They first evaluate performance of large and small models, showing that smaller base models underperform before fine-tuning (<a class=\"medius-aisummary\" data-starttime=\"00:16:12\">00:16:12</a>). Using Foundry’s UI, Omkar initiates supervised fine-tuning on thousands of curated traces. After training, the fine-tuned GPT-4.1 mini model surpasses its larger teacher models in performance with significant cost savings. Alicia notes the preservation of model integrity through graders for task and intent resolution, proving that fine-tuning does not diminish base capabilities and achieves better, cheaper, and faster outcomes.</p>\n\n<p><strong>Reinforcement Learning for Smarter Agents:</strong> The next section (<a class=\"medius-aisummary\" data-starttime=\"00:21:03\">00:21:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>) transitions to reinforcement learning to overcome the limits of distillation. Alicia introduces the concept: instead of just copying a teacher, the model learns from its own graded responses. Using refund decision tasks, the model generates multiple rollouts per prompt, which are scored by an LLM-based grader. That feedback then shapes its reasoning and tool use. Omkar illustrates this using Foundry’s SDK, connecting real tool calls hosted on function apps. He explains how telemetry and metrics such as rewards and reasoning token usage track learning progress and avoid reward hacking failures. Reinforcement fine-tuning results in noticeable improvement—raising accuracy from 74% to 84%—demonstrating how iterative grading and real tool use guide the model toward smarter decision-making while retaining cost benefits.</p>\n\n<p><strong>Interactive Low-Level Training API:</strong> Building on these techniques, Alicia introduces Foundry’s new interactive training API (<a class=\"medius-aisummary\" data-starttime=\"00:33:03\">00:33:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:06\">00:41:06</a>), designed for advanced practitioners who want more control. Unlike managed fine-tuning, users can define rollouts, losses, and update strategies while Microsoft handles GPU infrastructure. Omkar demonstrates using the API to fine-tune an open-source model (Qwen3 32B), employing primitives like forward and backward passes, sampling, and syncing checkpoints. He shows fine-grained telemetry—entropy, gradient norms, and KL divergence—that enables precise adjustment of algorithms such as PPO, DPO, or GRPO. Results reveal that even smaller open models trained via low-level control surpass earlier fine-tuned versions. Alicia highlights the advanced visibility this provides, allowing debugging, curriculum learning, and adaptive recipes securely within Foundry’s environment.</p>\n\n<p><strong>Simplifying Fine-Tuning and Closing Remarks:</strong> Near the end (<a class=\"medius-aisummary\" data-starttime=\"00:41:41\">00:41:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:47:06\">00:47:06</a>), Alicia introduces the fine-tuning skill integrated with GitHub Copilot for Azure, enabling users to execute fine-tuning tasks simply by describing their goals in natural language. The agent autonomously builds graders, extracts traces, filters data, and runs distillation experiments via autopilot. She demonstrates improved scores for fine-tuned smaller models, reiterating that fine-tuning now is accessible to everyone without requiring SDK or UI expertise. The talk concludes by dispelling myths: frontier models are not always the best because smaller fine-tuned ones can outperform them economically, and training costs are minimal with Foundry’s developer tiers. Participants are encouraged to explore public sample notebooks and sign up for previews or hands-on reinforcement learning labs. The speakers thank the audience, closing with an invitation to continue exploring model customization through Microsoft Foundry.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"22:45 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","speakerIds":["1768245294609build26-1772039473777001dTM9","1768245294609build26-1774284998169001pIgG","1768245294609build26-1769014422848001RlxU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.851Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/2e2fd0b4-5960-4c31-b9d2-946486fbbe5f","ogTitle":"Deploy. Observe. Learn. Reinforcement learning for production agents","ogDescription":"Agents don't fail in demos — they fail in production. In this breakout, see how teams use fine-tuning and reinforcement learning on Microsoft Foundry to improve production agents using real usage signals. We cover when fine-tuning reduces cost and latency, when RL delivers deeper gains, and how Foundry makes it easy to train, evaluate, and redeploy safely. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1769014422848001RlxU","RegistrantKey":"1768245294609build26-1769014422848001RlxU"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998169001pIgG","RegistrantKey":"1768245294609build26-1774284998169001pIgG"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK231","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"a21ba65c-ae2d-4261-a398-b164970d8261","startDateTime":"2026-06-02T22:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mark Russinovich, Ion Stoica, Darby Kosten","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018705273001rbGZ","langLocale":"en-US","sessionId":"1774018705273001rbGZ","sessionInstanceId":"1774018705273001rbGZ","sessionCode":"BRK227","sessionCodeNormalized":"EN-US-BRK227","title":"Distributed systems to AI platforms with Mark Russinovich & Ion Stoica","sortTitle":"distributed systems to ai platforms with mark russinovich & ion stoica","description":"What will it take to build AI platforms for the agent era? Join Azure CTO Mark Russinovich and UC Berkeley professor Ion Stoica (co-creator of Apache Spark and co-founder of Databricks and Anyscale) to explore how AI infrastructure must evolve as systems become agentic, multimodal, and globally distributed. Get practical insights on next generation architectures, from training to real time serving, and why open source, security, and governance are now core platform concerns.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"22:45 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/e28400c7-e07f-4eee-852d-345171931f8d","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e28400c7-e07f-4eee-852d-345171931f8d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e28400c7-e07f-4eee-852d-345171931f8d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e28400c7-e07f-4eee-852d-345171931f8d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e28400c7-e07f-4eee-852d-345171931f8d","speakerIds":["1768245294609build26-1776631798007001nGQM","1768245294609build26-1775513319814001teD5","1768245294609build26-1774976696107001NbL1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.462Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e28400c7-e07f-4eee-852d-345171931f8d","ogTitle":"Distributed systems to AI platforms with Mark Russinovich & Ion Stoica","ogDescription":"What will it take to build AI platforms for the agent era? Join Azure CTO Mark Russinovich and UC Berkeley professor Ion Stoica (co-creator of Apache Spark and co-founder of Databricks and Anyscale) to explore how AI infrastructure must evolve as systems become agentic, multimodal, and globally distributed. Get practical insights on next generation architectures, from training to real time serving, and why open source, security, and governance are now core platform concerns.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775513319814001teD5","RegistrantKey":"1768245294609build26-1775513319814001teD5"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774976696107001NbL1","RegistrantKey":"1768245294609build26-1774976696107001NbL1"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1776631798007001nGQM","RegistrantKey":"1768245294609build26-1776631798007001nGQM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK227","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Next.js","logicalValue":"Next.js"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Post Quantum Cryptography","logicalValue":"Post Quantum Cryptography"}]},{"@search.score":1.0,"documentId":"f5ad3414-7779-495b-9944-e3800060eb26","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Simon Willison","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709606001rFc1","langLocale":"en-US","sessionId":"1774018709606001rFc1","sessionInstanceId":"1774018709606001rFc1","sessionCode":"BRK208","sessionCodeNormalized":"EN-US-BRK208","title":"Ensuring your code works when AI testing isn’t enough","sortTitle":"ensuring your code works when ai testing isn’t enough","description":"Technical talk about building AI systems that scale for humans and machines, ensuring reliable code, and scalable systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Evolution of Agentic Engineering:</strong> The talk opens with Simon Willison introducing himself and outlining the theme of “verification patterns for agentic engineering” <a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>. He recounts his early experiences using LLMs like GPT‑3 for coding tasks as far back as 2022, describing how what began as “vibe coding” has transitioned into a structured discipline <a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>. Simon reflects on the emergence of “agentic engineering” — a professionalized form of software creation with autonomous code‑writing and testing agents — marking a shift that reached maturity after the “November inflection point” of 2025, when new coding‑focused LLMs made agent collaboration practical <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>.</p>\n\n<p><strong>The Rise of Verification and the Software Factory Concept:</strong> As the field matured, Simon stresses the importance of balancing speed and quality. He compares modern code productivity to traditional output levels, noting that while developers can now generate hundreds of lines of tested code per day, ensuring reliability remains the challenge <a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>. To illustrate the risks and innovations, he discusses examples like StrongDM’s “dark factory” — a system where software is produced, tested, and validated entirely by autonomous agents, following rules such as “code must not be written or reviewed by humans” <a class=\"medius-aisummary\" data-starttime=\"00:08:04\">00:08:04</a>. He highlights their use of scenario testing and “digital twin” environments that replicate APIs like Slack and Okta, allowing massive automated verification without rate limits <a class=\"medius-aisummary\" data-starttime=\"00:09:19\">00:09:19</a>.</p>\n\n<p><strong>Practical Verification Patterns and Code Review Methods:</strong> Moving into replicable strategies, Simon presents pragmatic techniques to bring verification discipline to smaller projects. He introduces “active refactoring,” a pattern where the developer aggressively edits and clarifies AI‑generated code instead of performing passive reviews <a class=\"medius-aisummary\" data-starttime=\"00:11:48\">00:11:48</a>. This approach, impossible in human collaboration, works well with agents since “you can’t be rude to an agent.” He also explores fast validation loops through continuous integration, test automation, and agentic test generation. By prioritizing high‑stakes code (for example, authentication flows) for manual inspection while automating low‑risk sections, developers can maintain speed without losing conceptual integrity <a class=\"medius-aisummary\" data-starttime=\"00:13:14\">00:13:14</a>.</p>\n\n<p><strong>Prototyping, Documentation, and Continuous Deployment:</strong> Simon explains that prototyping — once time‑consuming — is now near‑free thanks to agents that can instantly generate functional proofs of concept <a class=\"medius-aisummary\" data-starttime=\"00:15:14\">00:15:14</a>. He describes building his own agent prototypes and using them later as verified components for production systems. The newly emergent practice of “agentic documentation,” where LLMs generate concise and accurate code documentation, helps ensure long‑term trustworthiness by continuously checking diffs to keep docs up to date <a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>. He also emphasizes continuous deployment and preview environments as vital verification tools that allow engineers to explore live changes safely and improve collaboration <a class=\"medius-aisummary\" data-starttime=\"00:20:10\">00:20:10</a>.</p>\n\n<p><strong>Verification Resilience, Sandboxing, and Error Elimination:</strong> Later sections delve into isolation strategies to reduce the “blast radius” of software errors. Simon discusses practical sandboxing tools — such as CSP headers, isolated iframes, and WebAssembly (WASI) runtimes — that can safely execute untrusted code <a class=\"medius-aisummary\" data-starttime=\"00:22:20\">00:22:20</a>. He recounts experiments where agents stress‑tested these sandboxes to validate their security boundaries <a class=\"medius-aisummary\" data-starttime=\"00:23:43\">00:23:43</a>. He also champions a zero‑tolerance stance toward flaky tests, demonstrating how agents can now replicate complex CI environments and debug instability faster than humans <a class=\"medius-aisummary\" data-starttime=\"00:25:25\">00:25:25</a>. Through these methods, automated systems gain reliability comparable to large, multi‑layered industry QA pipelines, but scaled down for personal and small‑team workflows.</p>\n\n<p><strong>Concluding Reflections and Audience Discussion:</strong> In the closing discussion, Simon situates agentic engineering as both a technological and cultural evolution in software practice. He asserts that the same methods once viewed as enterprise luxuries — automated testing, continuous deployment, structured documentation — now serve as daily tools for individual developers because of agent acceleration <a class=\"medius-aisummary\" data-starttime=\"00:27:21\">00:27:21</a>. Drawing on historical texts like “The Mythical Man‑Month,” he notes that the principles of conceptual integrity and disciplined design apply more than ever. During Q&A <a class=\"medius-aisummary\" data-starttime=\"00:29:24\">00:29:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:44:50\">00:44:50</a>, he discusses language choice, regulatory concerns, code readability, and the importance of preserving agent transcripts for transparency. He concludes by urging developers to treat agentic systems as “verification machines” — tools to build better software faster, with higher quality and maintainability than before, ensuring that automation enhances, rather than replaces, engineering craftsmanship.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"18:30 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","speakerIds":["1768245294609build26-1776365809006001iDqG"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.466Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/f3ef9971-9c7e-4883-acdc-d4c2bd4ca732","ogTitle":"Ensuring your code works when AI testing isn’t enough","ogDescription":"Technical talk about building AI systems that scale for humans and machines, ensuring reliable code, and scalable systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776365809006001iDqG","RegistrantKey":"1768245294609build26-1776365809006001iDqG"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"App Mod","logicalValue":"App Mod"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"1f4331a1-afce-4e24-a242-d90f42467435","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pablo Castro, Allison Sparrow","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701434001rYZQ","langLocale":"en-US","sessionId":"1774018701434001rYZQ","sessionInstanceId":"1774018701434001rYZQ","sessionCode":"BRK246","sessionCodeNormalized":"EN-US-BRK246","title":"Foundry IQ: Fuel agents with enterprise knowledge and agentic retrieval","sortTitle":"foundry iq: fuel agents with enterprise knowledge and agentic retrieval","description":"Foundry IQ is Microsoft's context engineering platform that enables agents to unlock knowledge from everywhere, with state-of-the-art agentic RAG and enterprise-grade security built in. In this technical deep dive, we share the latest developments in Foundry IQ and Azure AI Search, new knowledge sources, agentic retrieval enhancements, and present retrieval performance innovations and evals for our 3rd year in a row — all designed to help you ship higher-quality agents faster.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction to Foundry IQ:</strong> The session opens with Pablo from the Foundry IQ team welcoming viewers and explaining that the focus will be on how to connect AI agents and models to company data and applications using knowledge sources (<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>). He describes Foundry IQ as the platform that simplifies this connection and outlines three foundational goals: ease of use, versatility for complex scenarios, and top-tier AI quality across retrieval and ranking (<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>). Each principle ensures that developers can get started quickly, scale as their needs evolve, and maintain relevance and precision across data queries.</p>\n\n<p><strong>Demonstrating Easy Setup and Knowledge Base Creation:</strong> To illustrate usability, Pablo demonstrates creating a simple knowledge base within Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:01:14\">00:01:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>). He uploads movie files from Wikipedia, names the collection “Movies Wiki,” and connects it seamlessly to an MCP server within Foundry IQ. Instead of writing a new agent, he uses GitHub Copilot to interact with this data by setting up a proxy that ties Foundry IQ services to Copilot through API endpoints (<a class=\"medius-aisummary\" data-starttime=\"00:03:10\">00:03:10</a>). After configuration, the demo shows the system answering queries—like identifying which pill Neo took from *The Matrix*—proving how quickly data can be linked, indexed, and queried in a real-world scenario (<a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>).</p>\n\n<p><strong>Architecture and Serverless Foundry IQ Launch:</strong> Pablo then explains the layered structure of Foundry IQ, comprising the Foundry integration layer, a complete retrieval engine, and a robust backend built on Azure AI Search (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>). He transitions to the provisioning process and announces an important milestone—the public preview of the **serverless version of Foundry IQ** (<a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>). This new capability allows developers to create services in 10–20 seconds, with auto-scaling and pay-per-use flexibility. Serverless Foundry IQ maintains enterprise-grade ranking, vector search, and security, making it ideal for dynamic agent workflows and scalable production environments. Pablo demonstrates provisioning a serverless instance in Foundry within seconds, underscoring its zero-friction setup (<a class=\"medius-aisummary\" data-starttime=\"00:12:03\">00:12:03</a>).</p>\n\n<p><strong>Expanding Knowledge Sources and AI Integration:</strong> The next segment discusses how Foundry IQ integrates data from files, object stores, Fabric, and Microsoft 365 environments (<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>). Pablo introduces Azure Content Understanding for document parsing, optical character recognition, and complex layouts, improving index quality and search precision (<a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>). He also showcases connections to Work IQ (for Microsoft 365 content), Fabric IQ (for structured analytics and SQL-based data access), and Web IQ (for web-grounded retrieval). By combining these sources, agents can retrieve contextually relevant insights from different repositories—document-based, analytical, or public web sources. He further explains how Foundry IQ ensures robust access control through Entra and Purview integration, propagating document-level permissions and sensitivity labels into search workflows (<a class=\"medius-aisummary\" data-starttime=\"00:28:00\">00:28:00</a>).</p>\n\n<p><strong>Retrieval Quality, Efficiency, and Improvements:</strong> The presentation then examines how Foundry IQ’s retrieval process operates through agentic orchestration, iterative query planning, and synthesis steps that ensure accurate, grounded answers (<a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>). Pablo highlights the evolution into a second generation of agentic retrieval technology, citing measurable improvements in recall, accuracy, and completeness compared to previous systems or hybrid search alone (<a class=\"medius-aisummary\" data-starttime=\"00:32:32\">00:32:32</a>). He acknowledges that this sophistication slightly increases latency but dramatically enhances understanding. The system’s new semantic rankers, model-specific prompting, and caching optimization further reduce token usage while improving performance and cost efficiency. Metrics shown demonstrate that fewer tokens now produce better answers, reflecting the team’s ongoing focus on retrieval science and pragmatic AI efficiency (<a class=\"medius-aisummary\" data-starttime=\"00:35:00\">00:35:00</a>).</p>\n\n<p><strong>Deep Dive, Control, and Conclusion:</strong> To close, Pablo explores the backend within Azure Portal, showing how developers can inspect or adjust indexes, transformation pipelines, and quantization settings for performance tuning (<a class=\"medius-aisummary\" data-starttime=\"00:41:00\">00:41:00</a>). He compares the real-time speed between small indexes and massive datasets—such as all of English Wikipedia—and demonstrates instant sub-second search at scale for 22 million content chunks (<a class=\"medius-aisummary\" data-starttime=\"00:43:22\">00:43:22</a>). He concludes by reiterating that all features, including serverless deployment, advanced retrieval, and integrated data intelligence, are available for use immediately via Foundry, the Azure portal, or SDKs. Pablo ends by thanking the audience and directing them to further sessions exploring related IQ technologies (<a class=\"medius-aisummary\" data-starttime=\"00:44:51\">00:44:51</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"16:00 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","speakerIds":["1768245294609build26-1774553037071001kDWR","1768245294609build26-1774553096990001lHQu"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.456Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ae80fe5d-7c1e-4fc5-ab1c-a37839c0a9bd","ogTitle":"Foundry IQ: Fuel agents with enterprise knowledge and agentic retrieval","ogDescription":"Foundry IQ is Microsoft's context engineering platform that enables agents to unlock knowledge from everywhere, with state-of-the-art agentic RAG and enterprise-grade security built in. In this technical deep dive, we share the latest developments in Foundry IQ and Azure AI Search, new knowledge sources, agentic retrieval enhancements, and present retrieval performance innovations and evals for our 3rd year in a row — all designed to help you ship higher-quality agents faster.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1774553037071001kDWR","RegistrantKey":"1768245294609build26-1774553037071001kDWR"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774553096990001lHQu","RegistrantKey":"1768245294609build26-1774553096990001lHQu"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"},{"displayValue":"Evaluations","logicalValue":"Evaluations"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"a628c7d2-7922-415b-b82c-189c000f875a","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Evan Boyle, Cassidy Williams, Katie Liu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708754001rO1v","langLocale":"en-US","sessionId":"1774018708754001rO1v","sessionInstanceId":"1774018708754001rO1v","sessionCode":"BRK203","sessionCodeNormalized":"EN-US-BRK203","title":"From CLI to PR: Automating the path to merged code","sortTitle":"from cli to pr: automating the path to merged code","description":"Everyone talks about agents, but the real challenge is applying them to daily sprints. Moving beyond chat, we'll show how GitHub Copilot functions as an agentic partner in your workflow by live-coding a full cycle—from planning in the terminal to delegating work to the cloud and automating PR reviews. No high-level abstractions here. Just technical mechanics: context management, advanced features with Copilot CLI, and the patterns that make agentic workflows actually stick.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with Cassidy Williams and Evan Boyle greeting the audience and setting the stage for a discussion about the GitHub Command Line Interface (CLI) <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>. Cassidy, from GitHub’s Developer Advocacy team, and Evan, an engineering manager for the CLI team, explain that their talk will focus on modern code workflows, particularly how to use the CLI to migrate and merge code efficiently. They reference GitHub’s scale, noting the staggering statistic of hundreds of millions of commits occurring weekly <a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>. The hosts emphasize the challenge not only of writing and committing code but also of ensuring it lands effectively through merges, reviews, and automation. Setting this context prepares the audience for a deep dive into how the CLI integrates with GitHub Copilot and AI-driven developer tools.</p>\n\n<p><strong>AI Tools and the Power of the CLI:</strong> After introducing the broader landscape, Cassidy highlights the ascendancy of AI development tools, situating GitHub Copilot as central to this productivity transformation <a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a>. The CLI is described as a “headless” powerhouse—a core engine that underlies the GitHub Copilot app, SDKs, and third-party integrations. Evan expands on its modern features, including code generation, automations, and slash commands that allow users to swap between AI models like Anthropic’s Claude, OpenAI’s GPT, and Google’s Gemini via simple terminal interactions <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. Skill integration within the CLI allows developers to add specialized capabilities, such as accessibility checks or automated visual difference screenshots in pull requests <a class=\"medius-aisummary\" data-starttime=\"00:03:25\">00:03:25</a>. This segment underscores how the CLI evolves alongside rapid AI model improvements and can be tailored to individual project needs. Built-in agents, from general-purpose ones to dedicated review agents, enhance code quality—Evan even mentions that GitHub’s Copilot review agent is currently the third-largest contributor to the company’s codebase <a class=\"medius-aisummary\" data-starttime=\"00:04:19\">00:04:19</a>.</p>\n\n<p><strong>Modes and Features for Planning, Reviewing, and Testing:</strong> The speakers then move into an in-depth explanation of key CLI modes—particularly “plan mode,” which allows developers to map out complex feature implementations collaboratively with AI assistance <a class=\"medius-aisummary\" data-starttime=\"00:04:42\">00:04:42</a>. Plan mode helps build a blueprint complete with tasks, questions, and validation steps. Evan explains that this preparatory stage often takes a significant portion of his workflow, enabling high-confidence one-shot implementations through specification-driven processes <a class=\"medius-aisummary\" data-starttime=\"00:05:28\">00:05:28</a>. Examples like “slash review,” “slash new,” and “slash PR” are highlighted—commands that spawn agents, create new sessions, and manage pull requests. The recently added “auto-fix” capability can resolve CI failures automatically <a class=\"medius-aisummary\" data-starttime=\"00:08:15\">00:08:15</a>. Cassidy demonstrates “slash fleet,” which runs multiple agents or models in parallel for testing or planning efficiency, and “slash research,” leveraging GitHub’s Blackbird engine to conduct code-aware searches across open source and private repos with citation support <a class=\"medius-aisummary\" data-starttime=\"00:09:50\">00:09:50</a>. The “experimental” mode is portrayed as a space for early adopters to test upcoming features such as voice mode, which enables spoken interaction directly in the terminal <a class=\"medius-aisummary\" data-starttime=\"00:10:50\">00:10:50</a>. Together, these innovations represent an AI-assisted coding ecosystem attuned to both research and real-time collaboration.</p>\n\n<p><strong>New Interfaces, SDKs, and Integrations:</strong> The presentation transitions to demonstrations of tangible user experiences, such as the new terminal user interface (TUI), which brings a visually organized, tabbed environment into the command line <a class=\"medius-aisummary\" data-starttime=\"00:13:43\">00:13:43</a>. Cassidy also announces the CLI’s integration with JetBrains IDEs, allowing developers to harness Copilot agents without leaving their editors <a class=\"medius-aisummary\" data-starttime=\"00:14:24\">00:14:24</a>. The GitHub Copilot SDK—built on top of the CLI—is introduced as a foundation for creating custom agent-driven applications across browsers, office software, or chat systems. The presenters share creative examples: a browser shopping bot capable of autonomous navigation and a PowerPoint generator that compiles meeting decks automatically from a user’s weekly CLI activities <a class=\"medius-aisummary\" data-starttime=\"00:15:47\">00:15:47</a>. These cases amplify GitHub’s vision of embedding agentic workflows wherever developers want them. Cassidy and Evan further highlight community creativity—from gamified interfaces to pixel art shells—all leveraging the same SDK layer in languages like Go, Node, Rust, and Python <a class=\"medius-aisummary\" data-starttime=\"00:17:39\">00:17:39</a>.</p>\n\n<p><strong>Live Coding Demonstration with Audience Participation:</strong> In a live build session, the duo invite attendees to contribute ideas through GitHub issues on a shared repository (evanboyle/build2026-cli-live) <a class=\"medius-aisummary\" data-starttime=\"00:18:01\">00:18:01</a>. They walk through the CLI’s TUI to select issues interactively, using voice mode to query the system. Demonstrations include enabling “YOLO mode,” which grants full execution permissions for autonomous operation <a class=\"medius-aisummary\" data-starttime=\"00:21:07\">00:21:07</a>, and switching between plan and implementation as they prototype a playful idea called “Always-on-top Octocat.” As audience questions surface, they elaborate on integrating workflow tools like SpecKit with plan mode and clarify that voice mode uses local models for privacy <a class=\"medius-aisummary\" data-starttime=\"00:25:13\">00:25:13</a>. They later use “rubber duck” features to cross-evaluate AI outputs between multiple model families for accuracy. The demo concludes with building and iterating through local agents using plan, fleet, and review, helping the audience envision how different features synchronize within one developer loop <a class=\"medius-aisummary\" data-starttime=\"00:35:47\">00:35:47</a>.</p>\n\n<p><strong>Advanced Questions, Scaling, and Conclusion:</strong> The final segment addresses technical audience questions on model management, token budgeting, and working with legacy or large-scale codebases <a class=\"medius-aisummary\" data-starttime=\"00:41:13\">00:41:13</a>. Evan explains that in massive projects such as Office or Windows (tens of millions of lines), GitHub is building local indexing and vector-search capabilities into the CLI for efficiency <a class=\"medius-aisummary\" data-starttime=\"00:42:39\">00:42:39</a>. He emphasizes testing-centric development, noting that nearly 70% of CLI and app code is composed of tests to ensure reliable AI-driven refactoring. Cassidy and Evan close by summarizing how GitHub Copilot’s CLI functions as a foundation for scalable, agent-based development—bridging planning, execution, and automation under one system. They invite viewers to explore further through GitHub’s free “Copilot CLI for Beginners” course and continue experimenting with new features <a class=\"medius-aisummary\" data-starttime=\"00:46:38\">00:46:38</a>. The session concludes with thanks and encouragement to adopt the CLI as part of developers’ everyday AI workflows <a class=\"medius-aisummary\" data-starttime=\"00:46:47\">00:46:47</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"16:00 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/b2b9417e-2b22-4732-aaac-24df66a625f7","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b2b9417e-2b22-4732-aaac-24df66a625f7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b2b9417e-2b22-4732-aaac-24df66a625f7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b2b9417e-2b22-4732-aaac-24df66a625f7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b2b9417e-2b22-4732-aaac-24df66a625f7","speakerIds":["1768245294609build26-1776135903186001Orl7","1768245294609build26-1774284997876001pQKi","1768245294609build26-1774284986220001p4nX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.465Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b2b9417e-2b22-4732-aaac-24df66a625f7","ogTitle":"From CLI to PR: Automating the path to merged code","ogDescription":"Everyone talks about agents, but the real challenge is applying them to daily sprints. Moving beyond chat, we'll show how GitHub Copilot functions as an agentic partner in your workflow by live-coding a full cycle—from planning in the terminal to delegating work to the cloud and automating PR reviews. No high-level abstractions here. Just technical mechanics: context management, advanced features with Copilot CLI, and the patterns that make agentic workflows actually stick.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774284986220001p4nX","RegistrantKey":"1768245294609build26-1774284986220001p4nX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284997876001pQKi","RegistrantKey":"1768245294609build26-1774284997876001pQKi"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776135903186001Orl7","RegistrantKey":"1768245294609build26-1776135903186001Orl7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK203","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"11f2d250-3e21-459f-beeb-8a53c021c3eb","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vivek Bhadauria, Sebastian Kohlmeier, Filisha Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018716883001rHBl","langLocale":"en-US","sessionId":"1774018716883001rHBl","sessionInstanceId":"1774018716883001rHBl","sessionCode":"BRK252","sessionCodeNormalized":"EN-US-BRK252","title":"From observability to ROI for AI agents on any framework","sortTitle":"from observability to roi for ai agents on any framework","description":"Nondeterministic, multi-agent systems break traditional monitoring. As agents reach production, observability must be built in—not added after failures. This session covers modern agent observability: cross-framework tracing and evals, rigorous inner-loop practices, evolving context-specific evals, and always-on signals that connect behavior to business outcomes to measure value, cost, and ROI.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Sebastian Kohlmeier welcomes viewers to the session \"From Observability to ROI for AI Agents on Any Framework.\" The talk introduces Microsoft Foundry Observability, a platform providing end-to-end tooling to monitor, evaluate, and optimize AI agents. Kohlmeier outlines the agenda, which includes key pillars—tracing, evaluation, monitoring, and optimization—and how these build into a full agent DevOps lifecycle. He describes the goal of enabling observability for any agent across frameworks to ensure reliability and quantify return on investment (<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>). Foundry’s observability features integrate across Microsoft’s AI ecosystem, unifying agent services, ML models, tools, and fine-tuning capabilities for both developers and operators.</p>\n\n<p><strong>Initial Demos and Rubric Evaluation:</strong> Filisha Shah joins at <a class=\"medius-aisummary\" data-starttime=\"00:03:55\">00:03:55</a> to demonstrate the “getting started” experience using an existing hosted agent that evaluates vendor performance in Microsoft data centers. She runs example queries, showing how Foundry’s tracing view captures live logs from agent interactions. As the session progresses, Filisha introduces the concept of evaluation as a way to measure agent reliability and introduces the new “rubric evaluator” feature at <a class=\"medius-aisummary\" data-starttime=\"00:05:53\">00:05:53</a>. This feature automatically creates multidimensional evaluation metrics based on a system prompt, model, and hosted agent target. The demo shows how Foundry generates rubrics across multiple weighted categories and applies them to production traces, helping diagnose agent errors, hallucinations, or incomplete task execution by examining low-performing traces and user prompts.</p>\n\n<p><strong>Open Ecosystem Support and Code‑First Observability:</strong> Following the initial demo, Kohlmeier explains that Foundry Observability now supports a broad open ecosystem (<a class=\"medius-aisummary\" data-starttime=\"00:10:05\">00:10:05</a>), enabling full compatibility with frameworks such as LangChain, LangGraph, and OpenAI SDK. Foundry also integrates deeply with Azure Monitor to provide full‑stack visualization from code to infrastructure. The rubric evaluator enters public preview, offering automatic, context‑specific evaluation without pre‑existing datasets. Filisha then demonstrates “code‑first observability” within Visual Studio Code (<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>) using the Foundry MCP server and toolkit extensions. Within the IDE, developers can analyze evaluation results, detect weak metric dimensions, and implement suggested improvements. The system links code changes directly to monitored data, grounding optimization recommendations in trace results, making model refinement iterative, traceable, and actionable.</p>\n\n<p><strong>Advanced Evaluation and Optimization at Scale:</strong> The session progresses to new analytics and optimization tools. Kohlmeier highlights multi‑turn evaluation and “user simulation” introduced at <a class=\"medius-aisummary\" data-starttime=\"00:17:59\">00:17:59</a>, which allow realistic test conversations and training in the absence of data. Foundry now enables feedback loops where traces feed back into training datasets to continuously expand coverage. Vivek Bhadauria then demonstrates “AZD AI Agent Optimize” (<a class=\"medius-aisummary\" data-starttime=\"00:20:23\">00:20:23</a>), showing how it autonomously improves an agent’s prompt, tool definitions, and models. The optimizer deploys and evaluates new agent variants, compares performance using rubric‑based metrics, and automatically converges to better configurations, raising accuracy by 27% in the demo. The system identifies areas such as hallucination issues and contextual grounding errors, refining system prompts and tool descriptions automatically. These optimization workflows support ongoing “hill‑climbing” on live trace data, aligning Foundry’s DevOps approach with AI’s iterative nature.</p>\n\n<p><strong>Agent ROI Analysis and Business Impact:</strong> In the final part (<a class=\"medius-aisummary\" data-starttime=\"00:27:19\">00:27:19</a>), Kohlmeier introduces the new \"Agent ROI\" module, now in private preview. This capability measures economic efficiency by pairing evaluator data with cost figures—token usage, tool expenses, and estimated labor savings. By assigning a business value (e.g., five dollars saved per successful agent outcome), users can track net value and ROI trends across agent versions. The feature compares LLM and tool costs and allows trace‑level investigation of low‑ROI executions to root cause inefficiencies using Foundry’s code‑first toolkit. Together, these analytics and optimization tools allow developers and stakeholders to demonstrate tangible returns from agentic workloads while maintaining reliability and performance visibility.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> The session closes at <a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a> with a recap of Foundry Observability’s unified workflow—from tracing and evaluation to monitoring, optimization, and ROI proof. Kohlmeier emphasizes Microsoft’s contribution to open telemetry standards, ongoing improvements in memory tracing semantics, and integration with red‑teaming, CI/CD pipelines, and Azure Monitor. He notes that customers like NTT DATA are already leveraging these capabilities to bring AI agents into production environments confidently. The presenters invite attendees to experiment with the Foundry lab, available online after the event, and highlight upcoming Build sessions covering advanced observability topics. The talk concludes with an invitation for developers to begin building with Foundry across GitHub and M365 integration channels, completing the lifecycle from creation to deployment.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:00 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/bc75ad49-f354-49ee-a449-69cb12bcf5fb","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/bc75ad49-f354-49ee-a449-69cb12bcf5fb","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/bc75ad49-f354-49ee-a449-69cb12bcf5fb","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/bc75ad49-f354-49ee-a449-69cb12bcf5fb","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/bc75ad49-f354-49ee-a449-69cb12bcf5fb","speakerIds":["1768245294609build26-1778597318722001S00J","1768245294609build26-1775086360974001J4XY","1768245294609build26-1774284995366001pNmt"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.857Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/bc75ad49-f354-49ee-a449-69cb12bcf5fb","ogTitle":"From observability to ROI for AI agents on any framework","ogDescription":"Nondeterministic, multi-agent systems break traditional monitoring. As agents reach production, observability must be built in—not added after failures. This session covers modern agent observability: cross-framework tracing and evals, rigorous inner-loop practices, evolving context-specific evals, and always-on signals that connect behavior to business outcomes to measure value, cost, and ROI.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775086360974001J4XY","RegistrantKey":"1768245294609build26-1775086360974001J4XY"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995366001pNmt","RegistrantKey":"1768245294609build26-1774284995366001pNmt"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778597318722001S00J","RegistrantKey":"1768245294609build26-1778597318722001S00J"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK252","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}]},{"@search.score":1.0,"documentId":"f428e0e3-2f3d-4ef3-a4d6-e55c041c25d6","startDateTime":"2026-06-02T22:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Tina Schuchman, Jeff Hollan, Takuto Higuchi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018700647001rtkm","langLocale":"en-US","sessionId":"1774018700647001rtkm","sessionInstanceId":"1774018700647001rtkm","sessionCode":"BRK241","sessionCodeNormalized":"EN-US-BRK241","title":"From prototype to production: build and run agents at scale","sortTitle":"from prototype to production: build and run agents at scale","description":"AI agents are transforming how developers build software—but shipping production-grade agents demands more. This session walks through the end-to-end lifecycle of building AI agents with Foundry Agent Service and Microsoft Agent Framework. See how to go from local prototyping to enterprise-grade hosted deployment with identity, secure networking, evaluations, and lifecycle management. Learn how coding agents like GitHub Copilot integrate directly into the workflow.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Vision for AI Agents:</strong> The video begins with Tina Schackman and Jeff Holland introducing themselves and outlining the session’s purpose—showing how to build, deploy, and operate enterprise-level AI agents using Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>). Tina speaks on how we’ve entered a new era of AI agents where “building” is no longer the challenge, but reliable scaling and operation are (<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>). She emphasizes that today’s agents are not static bots but dynamic teammates that spawn new agents, acquire new skills, and continuously improve (<a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>). The core premise presented is that enterprise success now comes from deploying interconnected, long-running agents across every business function—from supply chain to support—forming an adaptive AI operating system powered through Microsoft Foundry and its integration ecosystem of Azure, GitHub, and Microsoft 365 (<a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>).</p>\n\n<p><strong>Scenario and Demo Overview:</strong> The session transitions into a practical demo scenario from Azure networking operations: a fiber cable outage near a data center triggers an autonomous agent system to resolve it quickly (<a class=\"medius-aisummary\" data-starttime=\"00:04:39\">00:04:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>). Tina explains that their goal is to show how such a “fiber outage response agent” is built. This agent receives sensor alerts, verifies reliability data from Fabric IQ, supplier details from Work IQ, document data via Foundry content understanding, and dispatches field reps while tracking progress through D365 and Teams (<a class=\"medius-aisummary\" data-starttime=\"00:05:32\">00:05:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:23\">00:06:23</a>). The demo will follow the lifecycle phases—Build, Deploy, and Operate—demonstrating each step within Foundry’s integrated platform and how continuous learning occurs during production workloads (<a class=\"medius-aisummary\" data-starttime=\"00:06:27\">00:06:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:43\">00:06:43</a>).</p>\n\n<p><strong>Building Local Agents with Foundry Toolkit:</strong> Jeff begins the live demo in the Build phase using Visual Studio Code combined with GitHub Copilot and Foundry Toolkit (<a class=\"medius-aisummary\" data-starttime=\"00:08:13\">00:08:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>). He shows how developers can start from sample scenarios or generate agents dynamically. The Foundry Toolkit now generally available helps developers integrate tracing, evaluations, and best practices directly in the coding environment (<a class=\"medius-aisummary\" data-starttime=\"00:09:54\">00:09:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:04\">00:10:04</a>). With the Microsoft Agent Framework in Python, Jeff demonstrates how harnesses securely execute shell commands and code within isolated runtime environments. He explains Foundry Toolbox as the unified endpoint for managing all tools, authenticating services like Foundry IQ, Fabric IQ, Teams, and document intelligence (<a class=\"medius-aisummary\" data-starttime=\"00:11:25\">00:11:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:33\">00:13:33</a>). He activates a local debug session, interacts with the agent through chat, and seamlessly adds voice capability—turning it into a voice-enabled agent that responds to field queries in real time (<a class=\"medius-aisummary\" data-starttime=\"00:16:07\">00:16:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:42\">00:17:42</a>). This section underlines how Foundry simplifies end-to-end local creation and testing.</p>\n\n<p><strong>Deploying and Hosting Enterprise Agents at Scale:</strong> Tina follows up announcing production-ready releases: Microsoft Agent Framework v1.0, Foundry Toolkit in VS Code, and upcoming Foundry hosted agents with secure sandbox isolation and long-running support (<a class=\"medius-aisummary\" data-starttime=\"00:18:02\">00:18:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:59\">00:19:59</a>). Jeff demonstrates deployment of those agents as autonomous “claw-like” systems using Foundry Routines—making them proactive operators rather than reactive bots (<a class=\"medius-aisummary\" data-starttime=\"00:21:06\">00:21:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:25\">00:22:25</a>). He shows how each hosted agent session runs in isolation yet maintains durable state via Microsoft Durable Task Scheduler and resumes after human approval, seamlessly continuing long-running investigations (<a class=\"medius-aisummary\" data-starttime=\"00:23:25\">00:23:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:56\">00:25:56</a>). These agents can be published directly into Teams and Microsoft 365 Copilot with their own identity and permissions, enabling enterprise users to interact natively through existing productivity ecosystems (<a class=\"medius-aisummary\" data-starttime=\"00:26:15\">00:26:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:06\">00:27:06</a>).</p>\n\n<p><strong>Operating and Optimizing Production Agents:</strong> Tina introduces the Operate phase—highlighting Foundry’s new tracing and evaluation capabilities for hosted agents (<a class=\"medius-aisummary\" data-starttime=\"00:29:55\">00:29:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:30:02\">00:30:02</a>). Jeff then demonstrates visibility and optimization workflows, including the “trace replay” view where developers can replay entire agent interactions to analyze model reasoning, tool usage, and token consumption (<a class=\"medius-aisummary\" data-starttime=\"00:30:14\">00:30:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:31:14\">00:31:14</a>). Foundry introduces Rubric, a LLM-powered evaluator generator that creates context-aware scoring criteria automatically based on agent roles such as voice-friendly responses (<a class=\"medius-aisummary\" data-starttime=\"00:33:24\">00:33:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:34:40\">00:34:40</a>). The demo continues with automatic agent optimization loops via the Azure Developer CLI, iterating on variables such as prompts, models, and tool configurations until improved candidate versions emerge (<a class=\"medius-aisummary\" data-starttime=\"00:35:25\">00:35:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:38:06\">00:38:06</a>). The improvements are quantifiable—like an 11% gain in evaluation performance—showing how Foundry continuously enhances agents through production feedback cycles. Tina closes this section describing procedural memory, enabling agents to learn across sessions without starting from scratch (<a class=\"medius-aisummary\" data-starttime=\"00:39:45\">00:39:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:40:11\">00:40:11</a>).</p>\n\n<p><strong>Conclusion and Industry Adoption:</strong> The final segment underscores Foundry’s enterprise maturity: over 80,000 customers already run production agents, with use cases from firms like Twilio and KPMG leveraging Foundry hosted agents for global operations and client engagement (<a class=\"medius-aisummary\" data-starttime=\"00:40:16\">00:40:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:16\">00:41:16</a>). Tina and Jeff then recap the demo journey—building a local fiber-outage agent, hosting it securely, publishing through Teams, and continuously optimizing its behavior in production (<a class=\"medius-aisummary\" data-starttime=\"00:41:22\">00:41:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:42:35\">00:42:35</a>). They reaffirm Microsoft Foundry’s core value: enabling developers to build seamlessly, deploy powerfully, and operate with trust on one unified enterprise platform. The presentation ends with encouragement to attend related Build sessions and explore the Foundry booth, inviting developers to experience firsthand how Foundry scales autonomous AI agents globally (<a class=\"medius-aisummary\" data-starttime=\"00:42:58\">00:42:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:21\">00:43:21</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"22:45 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/af92c987-5029-4a5f-8a59-4a3931941ce5","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/af92c987-5029-4a5f-8a59-4a3931941ce5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/af92c987-5029-4a5f-8a59-4a3931941ce5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/af92c987-5029-4a5f-8a59-4a3931941ce5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/af92c987-5029-4a5f-8a59-4a3931941ce5","speakerIds":["1768245294609build26-1775078152694001T9OH","1768245294609build26-1774284996168001pgz1","1768245294609build26-1752002123322001UIRz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.455Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/af92c987-5029-4a5f-8a59-4a3931941ce5","ogTitle":"From prototype to production: build and run agents at scale","ogDescription":"AI agents are transforming how developers build software—but shipping production-grade agents demands more. This session walks through the end-to-end lifecycle of building AI agents with Foundry Agent Service and Microsoft Agent Framework. See how to go from local prototyping to enterprise-grade hosted deployment with identity, secure networking, evaluations, and lifecycle management. Learn how coding agents like GitHub Copilot integrate directly into the workflow.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996168001pgz1","RegistrantKey":"1768245294609build26-1774284996168001pgz1"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1752002123322001UIRz","RegistrantKey":"1768245294609build26-1752002123322001UIRz"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775078152694001T9OH","RegistrantKey":"1768245294609build26-1775078152694001T9OH"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK241","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Evaluations","logicalValue":"Evaluations"},{"displayValue":"Enterprise","logicalValue":"Enterprise"},{"displayValue":"Identity","logicalValue":"Identity"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"ebd8678d-83c1-474a-9120-89ca7ed3269b","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Charles Feddersen, Abe Omorogbe, Eric Hudson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018704521001rjdj","langLocale":"en-US","sessionId":"1774018704521001rjdj","sessionInstanceId":"1774018704521001rjdj","sessionCode":"BRK223","sessionCodeNormalized":"EN-US-BRK223","title":"From rows to reasoning: Designing databases for AI apps and agents","sortTitle":"from rows to reasoning: designing databases for ai apps and agents","description":"AI applications and agents require data platforms designed for reasoning, not just transactions. Traditional architectures force developers to stitch data systems together, adding latency and complexity. In this demo‑rich session, we’ll show the latest innovations in SQL Database and Cosmos DB, then build an app on Azure HorizonDB, Azure’s new cloud‑native PostgreSQL service, to show how AI apps built directly in the database simplifies design and enables reasoning over operational data.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:30 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","speakerIds":["1768245294609build26-1765362540584001w9jd","1768245294609build26-1765837379511001Rrqc","1768245294609build26-1773280139867001J6Je"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b3997e7a-32c9-4e3c-b10a-0f8012d0007e","ogTitle":"From rows to reasoning: Designing databases for AI apps and agents","ogDescription":"AI applications and agents require data platforms designed for reasoning, not just transactions. Traditional architectures force developers to stitch data systems together, adding latency and complexity. In this demo‑rich session, we’ll show the latest innovations in SQL Database and Cosmos DB, then build an app on Azure HorizonDB, Azure’s new cloud‑native PostgreSQL service, to show how AI apps built directly in the database simplifies design and enables reasoning over operational data.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362540584001w9jd","RegistrantKey":"1768245294609build26-1765362540584001w9jd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765837379511001Rrqc","RegistrantKey":"1768245294609build26-1765837379511001Rrqc"},{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1773280139867001J6Je","RegistrantKey":"1768245294609build26-1773280139867001J6Je"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK223","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"1b39b06e-c941-425c-9ee4-9e61ee4d94fd","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Priyanka Sharma, Mario Toffia","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018706445001rlk5","langLocale":"en-US","sessionId":"1774018706445001rlk5","sessionInstanceId":"1774018706445001rlk5","sessionCode":"BRK229","sessionCodeNormalized":"EN-US-BRK229","title":"From Skeptic to Superpower: Real‑World AI Coding Workflows That Scale","sortTitle":"from skeptic to superpower: real‑world ai coding workflows that scale","description":"AI coding tools promise massive productivity gains, but which ones actually work and why? In this session, we’ll walk through how two engineers use different tools daily: one using Claude Code + Cursor, the other GitHub Copilot CLI. We’ll cover what initially didn’t work, skepticism we had, where each tool shines, and where they fall down. We’ll show what this looks like for a team of 20 engineers that delivers at a scale of 200. No hype, just realistic workflows, tradeoffs, and lessons learned.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Background:</strong> At the start of the presentation <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:15\">00:00:15</a>, Priyanka Sharma and Mario Toffia from Thiink introduce themselves and welcome the audience. Priyanka outlines her background with the Cloud Native Computing Foundation, connecting her previous experiences with Kubernetes adoption to current patterns in the enterprise uptake of generative AI. She describes Thiink as an energy management company powering over 10,000 commercial and residential buildings. Mario shares his journey from telecom and IoT to developing Thiink’s Energy OS—an architecture enabling large language models to perform autonomous tasks. He recounts how GitHub Copilot inspired the realization that AI would soon handle most routine coding tasks, setting the tone for their talk about AI-powered workflows and automation at Thiink.</p>\n\n<p><strong>Thiink’s Use of AI and Early Adoption Challenges:</strong> Continuing at <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>, Priyanka explains how Thiink embraced AI tools under a company-wide initiative to work “bigger, better, faster.” Despite initial hesitation due to their involvement in critical infrastructure, the staff gradually transitioned from cautious proofs of concept to more sophisticated workflows. She recalls early frustrations building with generative models but notes a breakthrough in February 2026 with Claude Opus 4.6, when building new website sections became rapid and intuitive. Across departments, AI began accelerating work—automating complex processes like request-for-quote generation from multi-day workflows down to minutes. The results saved the company hundreds of thousands annually, though she humorously admits generative AI also enabled her “OCD” perfectionism on site redesigns. These changes illustrated how AI tools transformed both workflow speed and creativity while surfacing new operational dependencies.</p>\n\n<p><strong>New Tech, Old Problems – Vendor Lock-in and Lessons Learned:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:10:25\">00:10:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:30\">00:12:30</a>, Priyanka recounts a major setback in April 2026 when Anthropic banned Thiink’s organization, halting all Claude-based operations. This crisis highlighted issues of vendor lock-in, data portability, and cost management—similar to familiar challenges in cloud-native adoption. The team survived by switching to backup tools like Cursor, albeit at higher cost. She reflects that losing access to core AI tools mirrored classic FinOps problems: budgeting for token pricing, dependency risks, and workflow fragmentation. Priyanka distills key takeaways—avoid relying on tools without clear data export paths, anticipate downtime, and remember that even “brilliant colleagues,” whether human or AI, can quit unexpectedly. These experiences positioned Thiink to approach generative AI with both enthusiasm and caution, applying cloud-era wisdom to new AI ecosystems.</p>\n\n<p><strong>Harnessing AI Drift and Building Thiink’s Autonomous Flow:</strong> Mario resumes at <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a> to dive into technical detail. He defines the concept of “drift”—how LLMs gradually misinterpret architectural intent due to inconsistent contextual sync, paralleling human misunderstanding in software teams. To prevent codebase decay, Thiink created the “Harness,” a structured workflow combining domain-driven design principles, immutable documentation, and specialized sub-agents that break project requirements into well-defined PRDs and features. Through iterative Socratic-style reviews, each agent undergoes QA synchronization, security audits, and architectural alignment. Mario’s tools leverage multiple models (Claude 4.8 for implementation, GPT‑5.5 for QA) to cross‑validate tasks and maintain robust reliability. The Harness formalizes context management so AI workflows stay aligned with human architectural standards while automatically adapting code to feature dependencies, ensuring quality and traceability in self‑coded environments.</p>\n\n<p><strong>Live Demonstration and Workflow in Action:</strong> During <a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:34:00\">00:34:00</a>, Mario presents a live demo showing Thiink’s AI agents autonomously executing build tasks, performing smoke tests, and submitting pull requests. Priyanka excitedly notes the real-time progress as adversarial reviewers and automated gating pass successfully, culminating in a live merge on stage. Mario explains how the workflow autonomously distributes subtasks based on dependency graphs, performs adversarial code reviews across models, and handles version control while he monitors the results. This demonstration encapsulates their thesis: AI can handle significant components of production development while human engineers elevate focus to architecture, compliance, and design oversight. Mario concludes by describing his daily routine where AI executes overnight builds, freeing human teams to address critical decisions during work hours, particularly around sensitive security and ISO 27000 compliance issues.</p>\n\n<p><strong>Conclusion and Future Outlook:</strong> In closing <a class=\"medius-aisummary\" data-starttime=\"00:35:00\">00:35:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:38:00\">00:38:00</a>, Priyanka reflects that despite the excitement of generative AI’s era, familiar patterns re-emerge—portability, vendor neutrality, and FinOps concerns—echoing cloud-native adoption cycles. She emphasizes that while everyone can now “be a builder,” the democratization of AI development increases the surface area for potential drift and operational challenges. Thiink plans to remain nimble by backing open-source and vendor-neutral solutions, continuing to evolve their Harness and automation processes. The presenters celebrate their successful live merge as symbolic of collaboration between humans and AI, encouraging peers to embrace this “brilliant colleague” with curiosity and joy. They conclude with contact information and invitations to connect, underscoring that this moment marks not an ending but the beginning of a new, collaborative era of building with AI.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"16:00 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/23d1e046-18c1-49aa-ab56-9af87d6d4c31","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/23d1e046-18c1-49aa-ab56-9af87d6d4c31","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/23d1e046-18c1-49aa-ab56-9af87d6d4c31","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/23d1e046-18c1-49aa-ab56-9af87d6d4c31","speakerIds":["1768245294609build26-1776373731157001DADu","1768245294609build26-1776371578196001HCug"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.231Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/23d1e046-18c1-49aa-ab56-9af87d6d4c31","ogTitle":"From Skeptic to Superpower: Real‑World AI Coding Workflows That Scale","ogDescription":"AI coding tools promise massive productivity gains, but which ones actually work and why? In this session, we’ll walk through how two engineers use different tools daily: one using Claude Code + Cursor, the other GitHub Copilot CLI. We’ll cover what initially didn’t work, skepticism we had, where each tool shines, and where they fall down. We’ll show what this looks like for a team of 20 engineers that delivers at a scale of 200. No hype, just realistic workflows, tradeoffs, and lessons learned.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776373731157001DADu","RegistrantKey":"1768245294609build26-1776373731157001DADu"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776371578196001HCug","RegistrantKey":"1768245294609build26-1776371578196001HCug"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Windows APIs","logicalValue":"Windows APIs"}]},{"@search.score":1.0,"documentId":"d403abce-9e2d-407a-8e4b-7daa5f9527e5","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Tim Bozarth","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776966893131001OtRc","langLocale":"en-US","sessionId":"1776966893131001OtRc","sessionInstanceId":"1776966893131001OtRc","sessionCode":"BRK210","sessionCodeNormalized":"EN-US-BRK210","title":"Future of Developer Productivity: Microsoft’s EngThrive Framework in Practice","sortTitle":"future of developer productivity: microsoft’s engthrive framework in practice","description":"AI is transforming how we write code, but the teams shipping fastest go beyond tools—they systematically remove bottlenecks in the development process. This session gives you a practical framework to measure and improve developer productivity in the AI era, based on Microsoft’s EngThrive model of Speed, Ease, and Quality. See how to apply each dimension through case studies, including a team that reduced developer toil by 25%. You’ll leave with clear metrics to track, dashboards to build, and actions that work for teams of any size.","aiDescription":"<p><strong>Introduction and Focus:</strong> The talk begins with the speaker welcoming the audience and setting the stage for a discussion on productivity, specifically engineering productivity, in the age of AI (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>). The focus is on how AI can transform the way people work rather than just the products they build. The speaker emphasizes that while AI has two major value propositions—enhancing products and improving human processes—this session concentrates on the latter. They explain the structure of the session, noting that it will include contextual trends in the changing nature of work and a deep dive into Microsoft’s methodology for optimizing productivity, ending with an open Q&A session.</p>\n\n<p><strong>Modern Productivity and the Role of AI:</strong> The speaker highlights that traditional metrics of software development, such as lines of code or number of pull requests, are obsolete and even harmful if used as measures of individual performance (<a class=\"medius-aisummary\" data-starttime=\"00:03:15\">00:03:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>). At Microsoft, longitudinal studies divide developer time into innovation, operations, and organizational responsibilities, revealing that less than 15% of engineers’ work week is spent actively coding (<a class=\"medius-aisummary\" data-starttime=\"00:04:14\">00:04:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:22\">00:05:22</a>). As AI reshapes these distributions, the goal becomes “creating more value faster.” The development lifecycle’s shape is shifting; work previously concentrated on coding and system maintenance is now increasingly automated. Code is turning from an input to an output of AI systems, leading to an identity shift for engineers and new definitions of value creation (<a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>).</p>\n\n<p><strong>Microsoft’s Approach with EngeThrive:</strong> The core framework discussed is Microsoft’s “EngeThrive,” which focuses on making it fast and easy to build great work (<a class=\"medius-aisummary\" data-starttime=\"00:13:01\">00:13:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>). This philosophy breaks productivity into three measurable pillars: speed, ease, and quality. EngeThrive emphasizes measuring system-level productivity instead of individual output, assessing the responsiveness of environments and workflows rather than personal metrics. The speaker distinguishes between activity metrics—such as token usage or PR counts—and outcome metrics that describe real organizational value. Key outcome metrics include time from idea to customer delivery, the innovation time ratio, and measures of quality like defect escape and resilience (<a class=\"medius-aisummary\" data-starttime=\"00:19:29\">00:19:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:59\">00:20:59</a>). Consistency across company-wide metrics is highlighted as essential to meaningful improvement programs.</p>\n\n<p><strong>Program Implementation and Continuous Improvement:</strong> To operationalize EngeThrive, Microsoft built a leadership model emphasizing accountability, consistent measurement, and continuous improvement rather than static benchmarking (<a class=\"medius-aisummary\" data-starttime=\"00:21:43\">00:21:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:21\">00:23:21</a>). The framework promotes the identification and removal of bottlenecks in the software development lifecycle while empowering small teams to make measurable changes. Examples include using standard telemetry tools like Viva Insights to measure focus time and adopting dashboards and business rhythms to make metrics actionable. The speaker stresses that the real goal is iterative advancement, ensuring that teams develop a durable culture of improvement supported by data-informed decision-making rather than reactive reactions to arbitrary performance targets (<a class=\"medius-aisummary\" data-starttime=\"00:23:46\">00:23:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>).</p>\n\n<p><strong>Case Studies: Focus Time and Onboarding Speed:</strong> Two case studies demonstrate the framework in action. The first involved improving “focus time”—the uninterrupted blocks available for developers to enter a state of flow. By analyzing Viva Insights data and deploying better meeting discipline and calendar management, Microsoft increased average focus time by 2.1 hours per week per engineer, correlating with a 13% rise in PR velocity and recovery of 55,000 hours of engineering time within eight weeks (<a class=\"medius-aisummary\" data-starttime=\"00:25:27\">00:25:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:32\">00:28:32</a>). The second case study tackled onboarding efficiency, measured by “time to first pull request.” Through an AI onboarding assistant called First Mate, enhanced documentation, and short video exercises, new developers reduced their onboarding duration to under a week (<a class=\"medius-aisummary\" data-starttime=\"00:32:33\">00:32:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:35:06\">00:35:06</a>). This improvement translated empirically into faster progression to full productivity and better long-term retention—a testament to the power of outcome-driven measurement.</p>\n\n<p><strong>Conclusion and Q&A Insights:</strong> During the Q&A session (<a class=\"medius-aisummary\" data-starttime=\"00:37:08\">00:37:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:47:34\">00:47:34</a>), audience questions focused on measuring focus time, dealing with interruptions like meetings and messages, and convincing executives to adopt similar models. The speaker clarified that not all communications disrupt focus, and that revamping meeting culture—with clear agendas and permissions to decline—is key to preserving flow. They added that optimal focus sessions last around 120 minutes and reiterated that activity metrics lose their meaning once incentivized, whereas outcome-based metrics remain durable indicators of progress. The talk closes with a motivating reminder: even small, focused teams can drive organizational transformation by targeting systemic improvements, and AI now provides the means to accelerate meaningful change in engineering productivity.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"23:00 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/a10e37be-0b75-4535-8693-b49de03d99ac","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a10e37be-0b75-4535-8693-b49de03d99ac","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a10e37be-0b75-4535-8693-b49de03d99ac","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a10e37be-0b75-4535-8693-b49de03d99ac","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a10e37be-0b75-4535-8693-b49de03d99ac","speakerIds":["1768245294609build26-1778719549331001rAyH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.388Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a10e37be-0b75-4535-8693-b49de03d99ac","ogTitle":"Future of Developer Productivity: Microsoft’s EngThrive Framework in Practice","ogDescription":"AI is transforming how we write code, but the teams shipping fastest go beyond tools—they systematically remove bottlenecks in the development process. This session gives you a practical framework to measure and improve developer productivity in the AI era, based on Microsoft’s EngThrive model of Speed, Ease, and Quality. See how to apply each dimension through case studies, including a team that reduced developer toil by 25%. You’ll leave with clear metrics to track, dashboards to build, and actions that work for teams of any size.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778719549331001rAyH","RegistrantKey":"1768245294609build26-1778719549331001rAyH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"aef3699c-3c60-4002-995d-fa1c1217aca1","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nik Karpinsky, Mads Kristensen, Anisha Pindoria","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709436001rrGK","langLocale":"en-US","sessionId":"1774018709436001rrGK","sessionInstanceId":"1774018709436001rrGK","sessionCode":"BRK207","sessionCodeNormalized":"EN-US-BRK207","title":"GitHub Copilot in Visual Studio: Agents That Debug, Profile, and Test","sortTitle":"github copilot in visual studio: agents that debug, profile, and test","description":"Every IDE has agentic code generation now, but Visual Studio goes further with built-in agents that tackle what code generation can't. In this demo-heavy session, you'll see agents root-cause bugs using live runtime behavior, pinpoint performance bottlenecks and recommend targeted fixes, and build test coverage to catch regressions before they ship. These aren't just code-writing agents. They help enterprise C#, .NET, and C++ developers improve code quality with expert-level diagnostics.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Mads Kristensen and Nik Karpinsky from the Visual Studio team welcome attendees to their session at Build. Mads introduces the topic as Visual Studio and GitHub Copilot integration, clarifying that although Visual Studio has many features, this talk will focus specifically on Copilot and its roadmap. He sets the stage for understanding Visual Studio’s role in the new “agentic” development world where AI agents assist in coding. Mads emphasizes that Visual Studio remains the premier environment for professional C# and C++ developers on Windows, focusing on quality, governance, compliance, and code treated as an asset rather than a disposable artifact. This distinction defines how professional developers use Visual Studio—whether they author their own code or guide agents to do so while maintaining quality standards <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>.</p>\n\n<p><strong>Demo 1 – Debugging with Copilot and Agents:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>, Nik begins the technical demo inside the Visual Studio environment itself, working on a real bug fix within the Diagnostics Hub (internal profiler codebase). He shows how Copilot integrates with internal tools via MCP servers and the Azure DevOps (ADO) connector to fetch bug reports and create unit tests automatically. The demo showcases test-driven development using Copilot to reproduce a bug (“index out of range” in hybrid dictionary enumeration). Through iterative chat commands and confirmation prompts, the agent writes tests, identifies build failures, and suggests corrections in real time. Nik explains that Copilot now makes test-driven workflows intuitive—tests act as durable artifacts to validate changes and safeguard against regressions. Mads highlights the control and safety this approach offers professional developers <a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>.</p>\n\n<p><strong>Debugger Agent and Intelligent Fixes:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>, Nik transitions to the “debugger agent” mode in Copilot, demonstrating how the AI can set breakpoints, evaluate expressions, and reason about the code during debugging. The agent analyzes and resolves an off-by-one error detected within the hybrid dictionary collection, then verifies the fix by re-running tests. This segment illustrates the AI’s role as a collaborative assistant guiding through complex debugging scenarios rather than replacing developers. Mads and Nik note that understanding the system’s behavior remains critical—developers train and direct the AI, reinforcing the concept that good developers produce efficient AI outcomes. The demo shows successful resolution with the bug fixed before affecting end users, reflecting Copilot’s practical workflow benefits for enterprise-scale codebases like Visual Studio itself <a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>.</p>\n\n<p><strong>Demo 2 – Profiling Agent and Performance Optimization:</strong> Starting at <a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>, Nik demonstrates the new Visual Studio Profiler Agent that automatically captures performance traces during CI test runs using a VSTest performance collector. He explains how profiling traces record CPU and memory metrics, pinpointing bottlenecks. The profiler identifies inefficient “express stream read block” paths in a large file decoding test. Copilot analyzes actual runtime data—not just code heuristics—and generates optimized solutions to buffer reads and minimize overhead. After applying the suggested fix, Visual Studio rebuilds the solution and re-runs the profiler, confirming improved performance (reducing test time from 3.8s to 2.0s). Mads observes how integrated agents bring data-driven optimization into the daily developer “inner loop” rather than post-release profiling, with real measurable acceleration. They discuss applying similar profiling methods across Microsoft’s ecosystem, improving Visual Studio startup, Azure App Service footprint, and even Roslyn compiler efficiency <a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>.</p>\n\n<p><strong>Upcoming Features and Roadmap:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a>, Mads shifts to new and upcoming Copilot features for Visual Studio. He previews app modernization capabilities that can migrate legacy WebForms codebases to Blazor and .NET 10 using new automation agents. Additionally, the “agent skills” system will automatically suggest context-specific enhancements based on project types (e.g., WinForms or Azure) by integrating skills authored by internal product teams. Other enhancements include faster build initialization by avoiding redundant compile cycles when errors are detected, and AI-powered merge conflict resolution to intelligently handle version control merges that historically cause developer frustration <a class=\"medius-aisummary\" data-starttime=\"00:37:00\">00:37:00</a>. These developments improve Copilot’s integration reliability, user trust, and speed inside enterprise environments.</p>\n\n<p><strong>Infrastructure Advancements and Closing Highlights:</strong> Finally, from <a class=\"medius-aisummary\" data-starttime=\"00:39:00\">00:39:00</a> onward, Mads announces Visual Studio’s migration to the unified GitHub Copilot CLI SDK, ensuring future feature parity across VS Code, CLI, and Visual Studio. Visual Studio will retain unique deep integrations (like MSBuild and Roslyn APIs) but gain synced advancements as Copilot evolves. He further reveals future support for multi-model integration—allowing developers to connect any AI model, local or cloud-hosted, configured under enterprise security governance. This flexibility empowers organizations to choose trusted on-prem or external models while maintaining consistent AI toolchains. He concludes with encouragement to join the Insiders build next week for early access to these capabilities and thanks the audience for joining <a class=\"medius-aisummary\" data-starttime=\"00:43:00\">00:43:00</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"23:00 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/45d559ef-d829-4eda-a101-b034bea5bcb3","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/45d559ef-d829-4eda-a101-b034bea5bcb3","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/45d559ef-d829-4eda-a101-b034bea5bcb3","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/45d559ef-d829-4eda-a101-b034bea5bcb3","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/45d559ef-d829-4eda-a101-b034bea5bcb3","speakerIds":["1768245294609build26-1774645705844001UesE","1768245294609build26-1774284997266001pRH9","1768245294609build26-1774284992740001pOwG"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.466Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/45d559ef-d829-4eda-a101-b034bea5bcb3","ogTitle":"GitHub Copilot in Visual Studio: Agents That Debug, Profile, and Test","ogDescription":"Every IDE has agentic code generation now, but Visual Studio goes further with built-in agents that tackle what code generation can't. In this demo-heavy session, you'll see agents root-cause bugs using live runtime behavior, pinpoint performance bottlenecks and recommend targeted fixes, and build test coverage to catch regressions before they ship. These aren't just code-writing agents. They help enterprise C#, .NET, and C++ developers improve code quality with expert-level diagnostics.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"C++","logicalValue":"C++"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Content Lead","SpeakerId":"1768245294609build26-1774284992740001pOwG","RegistrantKey":"1768245294609build26-1774284992740001pOwG"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284997266001pRH9","RegistrantKey":"1768245294609build26-1774284997266001pRH9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774645705844001UesE","RegistrantKey":"1768245294609build26-1774645705844001UesE"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK207","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"3c506c8d-2865-4888-89df-7cf75c05ec99","startDateTime":"2026-06-03T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Simon Jakesch, Devanshi Joshi, Gopi Prashanth","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018704167001r1mq","langLocale":"en-US","sessionId":"1774018704167001r1mq","sessionInstanceId":"1774018704167001r1mq","sessionCode":"BRK221","sessionCodeNormalized":"EN-US-BRK221","title":"Idea to production-ready agent in seconds on AI-native runtime","sortTitle":"idea to production-ready agent in seconds on ai-native runtime","description":"Agentic apps behave differently from traditional services. They make decisions, coordinate steps, and react to changing inputs. This session shows how to run agentic workloads on a fast, AI native runtime on Azure Container Apps. The focus is on reliability and speed so you can move from idea to production‑ready agents. We’ll cover patterns for deployment, configuration, safe scaling for bursty traffic, and observability so teams can run agentic systems in production without surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Overview:</strong> The presentation opens with the host greeting the audience and setting the stage for the Microsoft Build event, emphasizing the energetic tone and inviting participation (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>). Product Marketing Manager Devanshi Joshi introduces herself and the session's goal: demonstrating how to put AI agents into production within seconds using Azure Container Apps on an AI-native runtime (<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:00\">01:00</a>). She is joined by Simon from the Microsoft product team and Gopi, a customer from Augur. The session outline highlights why agent deployments often fail, a hands-on demo of building from scratch, and how Azure Container Apps streamlines runtime decisions from infrastructure through optimization.</p>\n\n<p><strong>Runtime Challenges and Key Requirements:</strong> Devanshi discusses the common pitfalls in agentic AI projects, noting Gartner’s prediction that over 40% may be canceled by 2027 due to runtime—not model—issues (<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>). She outlines five critical runtime demands: cost control for runaway agent loops, security isolation to avoid untrusted code exposure, sub-second cold start performance, persistence for long-running agents, and seamless environment stitching (<a class=\"medius-aisummary\" data-starttime=\"00:02:52\">00:02:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:12\">00:07:12</a>). Each of these specifications represents a foundational expectation for modern agent runtimes, ensuring reliability across development and production. Her explanation sets the stage for Simon’s live demonstration of how Azure Container Apps delivers these runtime assurances at production scale.</p>\n\n<p><strong>Live Demonstration of Azure Container Apps and Sandboxes:</strong> Simon takes over to conduct a technical demo showcasing Azure Container Apps’ serverless scalability, sandboxing, and GPU flexibility (<a class=\"medius-aisummary\" data-starttime=\"00:07:51\">00:07:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:39\">00:08:39</a>). He shows multi-agent orchestration involving speech and text models, introducing three sandboxed agents—Aria, Nova, and an SRE agent—that communicate in real time via browser and voice connections. Through several interactions, he demonstrates Aria building a Tic Tac Toe game on port 80 (<a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:35\">00:14:35</a>), Nova retrieving meeting schedules and Teams messages, and the SRE agent monitoring infrastructure health. The demo validates cold start under 100 milliseconds, sandbox persistence, connector inheritance, and strong network isolation—key architectural features of Azure Container Apps sandboxes in public preview.</p>\n\n<p><strong>Technical Deep Dive and Feature Expansion:</strong> Simon further illustrates how sandboxes are organized into groups with connectors, volumes, and snapshots, demonstrating fast resume behavior and persistent memory capture (<a class=\"medius-aisummary\" data-starttime=\"00:17:22\">00:17:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:00\">00:23:00</a>). He quickly generates new sandboxes using preset templates, authenticates securely via GitHub tokens, and shows how Nova’s connectors inherit access to calendars and Teams without manual configuration. Persistent snapshots enable duplication and fast restoration of sandbox states for agent continuity. The sandbox experience concludes with a voice gateway connection example leveraging Twilio integrated with Azure Container Apps, linking AI agents to real-time communication scenarios (<a class=\"medius-aisummary\" data-starttime=\"00:24:00\">00:24:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>).</p>\n\n<p><strong>Product Announcements and Customer Use Cases:</strong> Transitioning back, Devanshi announces Azure Container Apps Sandboxes and Container Apps Express as a fast, isolated compute infrastructure executing workloads securely by default (<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:29:00\">00:29:00</a>). She explains how these capabilities support use cases across industries—from education technology environments to multi-tenant AI platforms—showcasing Sitecore AI and EdTech deployments. Gopi from Augur then presents how Augur is building autonomous supply chains using thousands of agents running on Azure Container Apps (<a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:40:30\">00:40:30</a>). He details Augur’s architecture: contextual agentic data integration across distributed systems, use of ACA to orchestrate real-time simulation pipelines, and augmentation through a unified world model called OSCO for scalable reasoning across enterprise supply chains.</p>\n\n<p><strong>Closing Insights and Q&A:</strong> Devanshi concludes the session summarizing that container apps and sandboxes are transforming AI agent production pipelines, now available in public preview (<a class=\"medius-aisummary\" data-starttime=\"00:41:12\">00:41:12</a>). She notes Microsoft’s own adoption of these runtimes for technologies like GitHub Copilot, Foundry, and other Azure services. Attendees are directed to official product blogs and repositories for hands-on resources and code access. The presentation ends with audience questions clarifying the distinction between Container Apps Express and Sandboxes, and Gopi elaborating on OSCO’s design as an object-oriented world model embedded in Augur’s codebase (<a class=\"medius-aisummary\" data-starttime=\"00:42:40\">00:42:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:10\">00:45:10</a>). The speakers thank participants and close the Build session on an enthusiastic note, inviting further experimentation with Azure’s agent-focused compute innovations.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:45 - 22:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/8508448f-e225-4efd-b0ae-f89c2c2b91b5","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8508448f-e225-4efd-b0ae-f89c2c2b91b5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8508448f-e225-4efd-b0ae-f89c2c2b91b5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8508448f-e225-4efd-b0ae-f89c2c2b91b5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8508448f-e225-4efd-b0ae-f89c2c2b91b5","speakerIds":["1768245294609build26-1776192489718001chCD","1768245294609build26-1750715119672001uIgP","1768245294609build26-1778674646909001gBFQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8508448f-e225-4efd-b0ae-f89c2c2b91b5","ogTitle":"Idea to production-ready agent in seconds on AI-native runtime","ogDescription":"Agentic apps behave differently from traditional services. They make decisions, coordinate steps, and react to changing inputs. This session shows how to run agentic workloads on a fast, AI native runtime on Azure Container Apps. The focus is on reliability and speed so you can move from idea to production‑ready agents. We’ll cover patterns for deployment, configuration, safe scaling for bursty traffic, and observability so teams can run agentic systems in production without surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1750715119672001uIgP","RegistrantKey":"1768245294609build26-1750715119672001uIgP"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776192489718001chCD","RegistrantKey":"1768245294609build26-1776192489718001chCD"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778674646909001gBFQ","RegistrantKey":"1768245294609build26-1778674646909001gBFQ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK221","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"App Mod","logicalValue":"App Mod"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Reserve","logicalValue":"Reserve"}]},{"@search.score":1.0,"documentId":"ad23e69c-3dbd-41ca-b899-a5bf9c050870","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mark Russinovich, Darby Kosten","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018705113001rDN4","langLocale":"en-US","sessionId":"1774018705113001rDN4","sessionInstanceId":"1774018705113001rDN4","sessionCode":"BRK226","sessionCodeNormalized":"EN-US-BRK226","title":"Inside Azure innovations with Mark Russinovich","sortTitle":"inside azure innovations with mark russinovich","description":"Join Mark Russinovich, CTO, Deputy CISO, Technical Fellow of Microsoft Azure. Mark will take you on a tour of the latest innovations in Azure architecture and explain how Azure enables intelligent, modern, and innovative applications at scale in the cloud, on‑premises, and on the edge.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The talk opens with Marcus, CTO and Deputy CISO for Azure, welcoming the audience to \"Inside Azure Innovations\" and setting expectations for the session (<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>). He explains that the presentation covers Azure’s latest infrastructure advancements, from data centers to cybersecurity, and emphasizes that some features shown may still be in development and might never ship as presented. The goal is to provide a behind-the-scenes look at Azure’s technology stack, connecting innovation across hardware and software to intelligence and AI systems (<a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>).</p>\n\n<p><strong>Infrastructure and Data Center Innovation:</strong> Marcus begins with Azure’s physical infrastructure, showcasing Fairwater data centers in Wisconsin and Atlanta (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:29\">00:02:29</a>). He gives a visual tour of the Atlanta site, under construction but already operational, highlighting its intricate cabling systems, liquid cooling, and server configurations. He recounts accidentally powering on a data cell during his visit, illustrating how active and iterative Azure’s engineering environment is. Azure Boost is then introduced as a central innovation—it uses offload cards to move data paths for storage and networking from CPUs to dedicated DPUs, boosting throughput significantly. Statistics show major improvements, including 400 Gbps networking and millions of IOPS for both local and remote storage (<a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:07\">00:05:07</a>).</p>\n\n<p><strong>Networking and Bare Metal Instances:</strong> The discussion shifts to Azure’s use of bare metal instances for GPU workloads, initially deployed for OpenAI at Fairwater (<a class=\"medius-aisummary\" data-starttime=\"00:05:36\">00:05:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:16\">00:06:16</a>). A demo shows SSH into a server proving that virtualized services are disabled, confirming true bare metal execution. Azure transitioned from InfiniBand to Ethernet-based MRC (Multi-Path Reliable Connection) networking, built collaboratively with OpenAI to make large multi-GPU training jobs resilient to congestion or switch failure. Marcus demonstrates how MRC maintains full 95 Gb/s throughput even after intentionally dropping one of four network switches (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:28\">00:09:28</a>). This leads to Remote Direct Memory Access (RDMA) over Azure Boost, which enables VM-to-VM memory transfers without the software networking stack, nearly doubling inference performance compared to TCP/IP, a key advancement for AI workloads (<a class=\"medius-aisummary\" data-starttime=\"00:13:08\">00:13:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:18\">00:13:18</a>).</p>\n\n<p><strong>Serverless and Container Virtualization:</strong> Marcus transitions to future infrastructure with serverless computing, emphasizing Azure Container Instances (ACI) as the foundation (<a class=\"medius-aisummary\" data-starttime=\"00:13:32\">00:13:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>). He explains ACI’s use of Hyper-V isolation instead of OS-process-level isolation for tighter security. The concept of direct virtualization—placing containers directly on servers as peers to management VMs—is introduced to remove overhead while keeping robust isolation (<a class=\"medius-aisummary\" data-starttime=\"00:16:19\">00:16:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:31\">00:16:31</a>). A live demo showcases container live migration, where a workload moves between servers without interruption or data loss (<a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:38\">00:18:38</a>). Container use now spans GitHub Actions, Python, Excel, and Horizon DB, with full fleet migration targeted over two years to direct virtualization for efficiency and resilience.</p>\n\n<p><strong>AI Systems and Context Optimization:</strong> Azure’s AI orchestration system, Manifold, is introduced to manage GPU clusters and enable complex multi-modal, multi-hardware workloads across CPU, GPU, and ASIC resources (<a class=\"medius-aisummary\" data-starttime=\"00:19:09\">00:19:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:45\">00:19:45</a>). A demo proves Manifold’s capability to deploy isolated pods with precise GPU and CPU allocation over direct virtualization. Marcus then presents Azure Context Cache, a new preview allowing token KV caches to be retained between inference sessions across servers using Azure Storage (<a class=\"medius-aisummary\" data-starttime=\"00:23:35\">00:23:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:26:02\">00:26:02</a>). This improves cache hit rates to over 96%, cutting cost and latency. Lastly, memory snapshotting accelerates container launches by reusing pre-initialized states, demonstrated by an astonishing live deployment of 10,000 sandboxes in just two seconds (<a class=\"medius-aisummary\" data-starttime=\"00:27:11\">00:27:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:35\">00:28:35</a>).</p>\n\n<p><strong>Security, Confidential Computing, and Future Hardware:</strong> Marcus closes with advances in data protection through confidential computing (<a class=\"medius-aisummary\" data-starttime=\"00:29:01\">00:29:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:53\">00:33:53</a>). He demonstrates the first-ever confidential VM live migration between servers, preserving attested trust boundaries. Further innovations include Azure Integrated HSMs—hardware security modules embedded in every server, maintaining keys within FIPS 140-3 level boundaries for full lifecycle protection. The demo shows encryption performance surging from 640 to 18,800 operations per second without reducing security (<a class=\"medius-aisummary\" data-starttime=\"00:37:09\">00:37:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:38:05\">00:38:05</a>). Finally, Marcus introduces Project Mosaic, an experimental optical interconnect technology from Microsoft Research Cambridge using micro-LED fibers for low-power, high-speed data transmission. A live demonstration displays individual LED modulation forming letters, proving the concept’s real-time responsiveness (<a class=\"medius-aisummary\" data-starttime=\"00:40:55\">00:40:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:55\">00:43:55</a>). Marcus concludes by summarizing innovations across all layers—data centers, serverless, AI, and hardware—and invites feedback before closing the session (<a class=\"medius-aisummary\" data-starttime=\"00:44:09\">00:44:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:44:46\">00:44:46</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/e959dc6a-bb39-46a9-84c5-b2620675b658","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e959dc6a-bb39-46a9-84c5-b2620675b658","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e959dc6a-bb39-46a9-84c5-b2620675b658","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e959dc6a-bb39-46a9-84c5-b2620675b658","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e959dc6a-bb39-46a9-84c5-b2620675b658","speakerIds":["1768245294609build26-1776631798007001nGQM","1768245294609build26-1774976696107001NbL1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e959dc6a-bb39-46a9-84c5-b2620675b658","ogTitle":"Inside Azure innovations with Mark Russinovich","ogDescription":"Join Mark Russinovich, CTO, Deputy CISO, Technical Fellow of Microsoft Azure. Mark will take you on a tour of the latest innovations in Azure architecture and explain how Azure enables intelligent, modern, and innovative applications at scale in the cloud, on‑premises, and on the edge.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Support In-person, Session Owner","SpeakerId":"1768245294609build26-1774976696107001NbL1","RegistrantKey":"1768245294609build26-1774976696107001NbL1"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1776631798007001nGQM","RegistrantKey":"1768245294609build26-1776631798007001nGQM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK226","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"a011ccc9-b119-4096-9d44-0abfb6f3e928","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Tadashi Okazaki","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776196765191001w7vt","langLocale":"en-US","sessionId":"1776196765191001w7vt","sessionInstanceId":"1776196765191001w7vt","sessionCode":"BRK209","sessionCodeNormalized":"EN-US-BRK209","title":"Japan Wrap-up Session","sortTitle":"japan wrap-up session","description":"本セッションは、日本からご参加の方向けに、Microsoft Buildの主要なアナウンスについて日本語で解説を行います。\n\nIn this session, we will deliver a focused recap of the key announcements and technical insights from the 2 days of Microsoft Build, tailored for developers joining from Japan. You will leave with a clear understanding of the technical direction revealed at Build, how it applies to your products and teams, and the concrete next steps to consider as you move toward frontier transformation.\n\nThis session will be delivered in Japanese and will include Japanese captions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Objectives:</strong> The session begins with greetings and an overview of the event, marking the start of the Japan Wrap-Up session for Microsoft Build 2026 (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>). The host explains that this 45-minute session will summarize the most important announcements from Build 2026 and present them clearly in Japanese for both livestream and local attendees (<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>). The goal is to provide attendees with concise insights and business takeaways, covering three key areas: cloud and AI, Windows and developer tools, and new frontier innovations (<a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>). Experts from different fields at Microsoft Japan are introduced, and the first speaker, Executive Officer Okazaki, takes the stage to discuss Build 2026’s primary themes and announcements.</p>\n\n<p><strong>Global Developer Focus and Infrastructure Innovation:</strong> Okazaki contextualizes Build 2026 within Microsoft’s broader mission to create an ecosystem powered by frontier AI and autonomous agents (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:33\">00:03:33</a>). He highlights the physical event’s festival-like atmosphere with 2,500 attendees onsite and about 150,000 online participants, underscoring global developer enthusiasm. Key innovations introduced by Satya Nadella and team focus on optimizing AI infrastructure performance and cost efficiency (“performance equals tokens”). Microsoft introduced new chipsets—Maia 200 and Cobalt 200—designed to enhance energy efficiency and AI workload processing across cloud and edge (<a class=\"medius-aisummary\" data-starttime=\"00:05:29\">00:05:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:04\">00:06:04</a>). Database breakthroughs, like Azure’s HorizonDB public preview and GPU-accelerated Fabric Data Warehouse, were showcased to strengthen enterprise-scale data operations (<a class=\"medius-aisummary\" data-starttime=\"00:06:45\">00:06:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:23\">00:07:23</a>). Additionally, Microsoft reinforced its Copilot and agent safety standards through technologies like the Microsoft Execution Container and Host Agent for secure, scalable deployment (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>).</p>\n\n<p><strong>Frontier Agents, Microsoft Scout, and AI Model Ecosystem:</strong> Moving into Microsoft’s frontier vision, Okazaki explains how autonomous systems, exemplified by Microsoft’s new Scout agent (<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:14\">00:13:14</a>), demonstrate practical automation across everyday tasks like business expense processing. Scout’s secure-human-in-the-loop design emphasizes safe execution while leveraging long-context frontier models that can process extended reasoning tasks with high accuracy (~80% correctness rate). Okazaki highlights advancements in model tuning (“frontier tuning”) that tailor AI models such as My Thinking King One to enterprise data and values, alongside collaboration with partners like Anthropic and Fireworks AI for model diversity. Microsoft’s multiverse of AI models now spans over 11,000 variants through the Microsoft AI Fund, covering text, vision, and reasoning capabilities (<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:28\">00:16:28</a>). These innovations aim to help organizations continuously refine agents’ intelligence and integrate AI securely into human workflows.</p>\n\n<p><strong>Windows, Local AI, and Developer Tools Evolution:</strong> The next segment by Okawa focuses on Windows and the local AI development environment (<a class=\"medius-aisummary\" data-starttime=\"00:22:41\">00:22:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:39\">00:24:39</a>). Microsoft positions Windows as the essential platform for both cloud and edge AI developers. New ARM-based Surface devices with GPU/NPU integration can now execute large language models locally (“foundry local AI”) for latency-sensitive or private applications (<a class=\"medius-aisummary\" data-starttime=\"00:24:01\">00:24:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:08\">00:25:08</a>). Updates such as Windows AI APIs, WebNN support for browser-based inference, and expanded GPU/CPU compatibility enable developers to accelerate AI workloads natively. Developer convenience was enhanced through WSL containers, streamlined environment setup via Dev Configuration, and the introduction of the powerful Windows 365 Cloud PC for scalable development (<a class=\"medius-aisummary\" data-starttime=\"00:26:41\">00:26:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:09\">00:28:09</a>). GitHub Copilot has also evolved—its new desktop app and Canvas collaboration feature allow multiple human and AI agents to co-develop, automate code reviews, and implement continuous integration pipelines through the Copilot SDK GA (<a class=\"medius-aisummary\" data-starttime=\"00:29:20\">00:29:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:26\">00:33:26</a>).</p>\n\n<p><strong>Cloud AI Platform, Observability, and Optimization Tools:</strong> Nitta introduces updates to the cloud-side agent platform. Key product releases include “IQ Serverless,” enabling serverless AI search with elastic scaling (<a class=\"medius-aisummary\" data-starttime=\"00:34:11\">00:34:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:34:52\">00:34:52</a>), and enhancements to Microsoft Execution Container (MXC), which now provides a secure runtime foundation for multi-agent applications. Observability tools were expanded through “Rubric Evaluator” and “Agent Optimizer,” which automatically assess conversational quality, continuously refine prompts, and measure cost-to-value efficiency via ROI Agent analytics (<a class=\"medius-aisummary\" data-starttime=\"00:38:04\">00:38:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:39:26\">00:39:26</a>). These capabilities help enterprises monitor and improve AI governance. Yamamoto later discusses infrastructure scaling, highlighting Microsoft’s AI data center expansion focused on sustainability (Fairwater project), modernization support through Azure migration Copilot for mainframe-inclusive architectures, and the unveiling of Layfin—a framework that allows rapid AI-powered backend construction for production-ready applications (<a class=\"medius-aisummary\" data-starttime=\"00:41:04\">00:41:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:03\">00:45:03</a>).</p>\n\n<p><strong>Closing Remarks and Community Engagement:</strong> The wrap-up concludes by encouraging attendees to share feedback via survey forms (<a class=\"medius-aisummary\" data-starttime=\"00:46:30\">00:46:30</a>), participate in follow-up networking at the Japan dinner event, and look forward to future Microsoft conferences like Ignite in November and next year’s Build (<a class=\"medius-aisummary\" data-starttime=\"00:46:40\">00:46:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:46:59\">00:46:59</a>). The presenters reaffirm Microsoft’s commitment to collaboration, innovation, and the AI-agent ecosystem, thanking the audience for their active participation and contributions throughout the session.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/52082998-f4df-4fbb-8b20-f285b89fd486","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/52082998-f4df-4fbb-8b20-f285b89fd486","captionFileLink":"","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/52082998-f4df-4fbb-8b20-f285b89fd486","speakerIds":["1768245294609build26-1770723178216001H10N"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.902Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/52082998-f4df-4fbb-8b20-f285b89fd486","ogTitle":"Japan Wrap-up Session","ogDescription":"本セッションは、日本からご参加の方向けに、Microsoft Buildの主要なアナウンスについて日本語で解説を行います。\n\nIn this session, we will deliver a focused recap of the key announcements and technical insights from the 2 days of Microsoft Build, tailored for developers joining from Japan. You will leave with a clear understanding of the technical direction revealed at Build, how it applies to your products and teams, and the concrete next steps to consider as you move toward frontier transformation.\n\nThis session will be delivered in Japanese and will include Japanese captions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770723178216001H10N","RegistrantKey":"1768245294609build26-1770723178216001H10N"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"27f0a016-e1b7-4348-bc3e-096734617036","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Michael Chiang, Parth Sareen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775517798562001H69f","langLocale":"en-US","sessionId":"1775517798562001H69f","sessionInstanceId":"1775517798562001H69f","sessionCode":"BRK235","sessionCodeNormalized":"EN-US-BRK235","title":"Local models, developer control, and the future of AI runtimes","sortTitle":"local models, developer control, and the future of ai runtimes","description":"How local and hybrid model execution is reshaping developer workflows, privacy, and experimentation. Why “run it yourself” is back. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session begins with Parth Sareen and Michael Chiang from Ollama greeting the audience and introducing themselves <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:10\">00:00:10</a>. They outline the focus of the talk—open models for intelligent agents and how hybrid local-cloud systems are enabling new applications <a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>. Michael briefly explains what Ollama is: a platform that allows developers to easily access, run, and integrate open models with their existing tools. He describes Ollama’s simplicity, where developers can execute models with one command, and introduces the \"ollama launch\" feature that connects open models with popular tools like VS Code, GitHub Copilot, and Claude Code <a class=\"medius-aisummary\" data-starttime=\"00:00:57\">00:00:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>. He highlights Ollama’s compatibility with major APIs (OpenAI and Anthropic) and SDKs for Python and JavaScript, emphasizing that the system was built from the ground up for agentic use, including structured outputs and one-line setup for agents.</p>\n\n<p><strong>Platform Growth and Hybrid Capabilities:</strong> Chiang continues with statistics showing Ollama’s adoption—over eight million active developers and partnerships with leading model labs and hardware firms <a class=\"medius-aisummary\" data-starttime=\"00:01:53\">00:01:53</a>. He mentions that newly announced models, such as Google DeepMind’s Gemma 4 12B Unified Model, are already available on Ollama <a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>. He introduces Ollama Cloud, launched earlier this year, which complements local inference by providing scalable, zero data retention processing on data-center hardware. This hybrid configuration allows users to run high-performance models remotely when local compute is constrained <a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:50\">00:02:50</a>. Chiang underscores how this setup enables advanced multi-agent workflows and tasks requiring large context windows while maintaining user data privacy. He also notes that open-source models are rapidly improving and now support mainstream applications at lower costs, offering flexibility for users to personalize models and parameters <a class=\"medius-aisummary\" data-starttime=\"00:03:19\">00:03:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>.</p>\n\n<p><strong>Evolution and Real-World Applications:</strong> Chiang details how open models have advanced—from simple chat functions to reasoning, multi-tool use, and long-horizon task execution spanning hours or days <a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:50\">00:04:50</a>. He then highlights real-world deployments of Ollama within research and enterprise environments. The Lawrence Berkeley National Laboratory uses it for autonomous physics research with an \"Accelerator Assistant\" that integrates cloud and local inference <a class=\"medius-aisummary\" data-starttime=\"00:05:13\">00:05:13</a>. NASA’s Glenn Research Center employs Ollama for probabilistic risk assessment, organizing Mars mission tasks into human-system categories <a class=\"medius-aisummary\" data-starttime=\"00:05:44\">00:05:44</a>. The U.S. Department of Energy and Brookhaven Lab use Ollama with LangChain to analyze logs and summarize results through natural language queries <a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:32\">00:06:32</a>. Beyond research, Ollama powers private and mission-critical tasks within enterprises, such as financial data reviews and CAD designs while keeping sensitive information local <a class=\"medius-aisummary\" data-starttime=\"00:06:42\">00:06:42</a>.</p>\n\n<p><strong>Demos and Hybrid Workflows:</strong> Parth Sareen transitions to live demonstrations <a class=\"medius-aisummary\" data-starttime=\"00:07:36\">00:07:36</a>. He showcases typical model-chatting workflows and illustrates how Ollama offers identical experiences across local and cloud deployments. Using \"ollama launch,\" he connects open models with different tooling systems like Copilot CLI and agentic frameworks such as OpenClaw and Hermes <a class=\"medius-aisummary\" data-starttime=\"00:08:17\">00:08:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>. Sareen demonstrates debugging tasks through Copilot, where subagents explore codebases and produce structured plans for developers <a class=\"medius-aisummary\" data-starttime=\"00:09:19\">00:09:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:33\">00:09:33</a>. He emphasizes hybrid execution, where cloud models handle complex workloads while local inference ensures privacy for sensitive tasks. A practical example involves locally analyzing a credit card statement using the Pi harness and a Qwen3.6 model entirely on-device, showing fast, private data processing without external transmission <a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:19\">00:12:19</a>.</p>\n\n<p><strong>Personal Agents and Ease of Setup:</strong> Sareen further demonstrates personal agent setups like OpenClaw, highlighting how users can install and launch agents end-to-end within seconds using \"ollama launch\" <a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:28\">00:13:28</a>. He explains the advantage of local model execution for privacy, allowing personalized intelligence tools to operate entirely on the user’s device. This section underscores Ollama’s goal of democratizing agentic capability by simplifying integration and minimizing setup time for productive, secure environments. The demo concludes with an invitation for audience questions about local and cloud model uses <a class=\"medius-aisummary\" data-starttime=\"00:13:41\">00:13:41</a>.</p>\n\n<p><strong>Q&A and Future Outlook:</strong> During the Q&A <a class=\"medius-aisummary\" data-starttime=\"00:14:04\">00:14:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>, audience members ask about edge and mobile integrations, VRAM, unified memory, and hardware compatibility. Chiang explains that while Ollama can run on Snapdragon-based devices, the focus remains on mainstream desktop use. Sareen and Chiang discuss quantization, unified RAM in Apple and NVIDIA chips, and Ollama’s MLX inference engine developed for Apple devices. Other attendees inquire about coding model recommendations for large-memory systems—Sareen mentions Qwen 3.6-27B and Gemma as top performers. He advises experimenting and benchmarking multiple models to find optimal fits for different workloads. The session closes with both presenters thanking the audience, encouraging deeper engagement, and inviting informal discussion afterward <a class=\"medius-aisummary\" data-starttime=\"00:20:05\">00:20:05</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0f29bea1-6955-4ec3-8c51-5a80b47dcf7b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0f29bea1-6955-4ec3-8c51-5a80b47dcf7b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0f29bea1-6955-4ec3-8c51-5a80b47dcf7b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0f29bea1-6955-4ec3-8c51-5a80b47dcf7b","speakerIds":["1768245294609build26-1772238023541001lKAd","1768245294609build26-1776980112222001zvph"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0f29bea1-6955-4ec3-8c51-5a80b47dcf7b","ogTitle":"Local models, developer control, and the future of AI runtimes","ogDescription":"How local and hybrid model execution is reshaping developer workflows, privacy, and experimentation. Why “run it yourself” is back. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772238023541001lKAd","RegistrantKey":"1768245294609build26-1772238023541001lKAd"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776980112222001zvph","RegistrantKey":"1768245294609build26-1776980112222001zvph"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Purview","logicalValue":"Purview"},{"displayValue":"Purview","logicalValue":"Purview"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"}]},{"@search.score":1.0,"documentId":"4a87a7a0-fd66-4041-a5e8-de82fc37ef64","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rochak Mittal, Adity Agarwal, Shobhit Garg","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018705660001rbvk","langLocale":"en-US","sessionId":"1774018705660001rbvk","sessionInstanceId":"1774018705660001rbvk","sessionCode":"BRK228","sessionCodeNormalized":"EN-US-BRK228","title":"Modern resiliency from build to recovery through Agentic AI","sortTitle":"modern resiliency from build to recovery through agentic ai","description":"Resiliency failures rarely come from missing features; they stem from fragmentation. During incidents teams lose time jumping between repos, dashboards, runbooks, and chat tools. This session reframes resiliency as an agent-first practice spanning build, operate, troubleshoot, and recovery. See how an agentic AI assistant links workflows across IDEs, source repos, and collaboration tools speeding validation, operations, and recovery for a critical workload aligned with Azure’s agentic model.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Core Principles:</strong> The session begins with Abhimanyu, joined by Aditi and Shobhit, introducing the theme of resiliency and new agent capabilities designed to help users identify and mitigate gaps in their systems (<a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a>). They define resiliency as the ability of systems to stay operational despite unexpected failures. Failures, they note, are inevitable—ranging from availability zone outages to regional failures and data loss. Azure offers foundational tools such as availability zones, geo-replication, and strong backup mechanisms to ensure continued operation. The discussion then segments resiliency into three main pillars: infrastructure, data, and cyber resiliency (<a class=\"medius-aisummary\" data-starttime=\"00:01:22\">00:01:22</a>), emphasizing that overall goals include minimizing downtime, preventing data loss, and enabling rapid recovery.</p>\n\n<p><strong>Resiliency Lifecycle and Azure Capabilities:</strong> The presenters outline the resiliency lifecycle, which encompasses the three stages of “Start Resilient,” “Get Resilient,” and “Stay Resilient.” Starting resilient means designing for continuity from day one using Azure's tools such as availability zones, region pairs, and the newly introduced Resiliency Agent within Azure Copilot (<a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>). Existing workloads can improve through the “Get Resilient” phase using automated gap detection and prioritization features from Azure Advisor and AI-powered experiences that surface critical reliability recommendations. Finally, the presenters describe “Stay Resilient” as an ongoing effort involving periodic drills, monitoring, and recovery planning to maintain readiness against real-world disruptions (<a class=\"medius-aisummary\" data-starttime=\"00:04:32\">00:04:32</a>).</p>\n\n<p><strong>Unified Resiliency Management and Public Preview Announcements:</strong> The speakers highlight the evolution from Azure Business Continuity Center to a unified “Azure Resiliency” platform announced at Ignite (<a class=\"medius-aisummary\" data-starttime=\"00:05:11\">00:05:11</a>). This solution integrates infrastructure, data, and cyber resiliency management into a single interface, enabling centralized visualization, proactive risk tracking, and simplified remediation. They introduce the concept of service groups, which shift planning from individual resources to entire applications, reflecting a business-centric resilience model. New additions include the Resiliency Manager public preview—providing zonal resiliency management, automated goal assignments, continuous health checks, and simulated resiliency drills powered by Chaos Studio (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>), allowing developers to test and improve their failover readiness through real-world failure simulations.</p>\n\n<p><strong>Demonstration: Building Apps to “Start Resilient”</strong> The “Start Resilient” demonstration shows how DevOps engineers can leverage the Resiliency Agent with Azure Copilot to generate resilient infrastructure-as-code templates directly from natural language instructions (<a class=\"medius-aisummary\" data-starttime=\"00:10:03\">00:10:03</a>). In the demo, the agent interprets a developer’s prompt to build a zonally resilient Java payroll application that includes a VM scale set, a PostgreSQL database, and Key Vault integration. It automatically generates Bicep templates, checks service compatibility with availability zones, and outputs deployment-ready files. The agent then directly integrates with GitHub to create pull requests, showcasing how resiliency can be embedded at code authoring time rather than retrofitted later. This workflow demonstrates how Azure Copilot’s AI-guided experience eliminates the need for manual configurations while enforcing best practices (<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>).</p>\n\n<p><strong>Retrofitting Existing Apps and Advanced Advisor Capabilities:</strong> The “Get Resilient” phase focuses on brownfield applications already deployed on Azure (<a class=\"medius-aisummary\" data-starttime=\"00:22:15\">00:22:15</a>). The demo of the “Ask HR” Java app shows how developers can group resources under a service group, assign zonal resiliency goals, and automatically assess posture compliance using the Resiliency Agent. Where gaps exist, the tool recommends and even applies configurations such as enabling high availability (HA) for PostgreSQL databases. Later, in Visual Studio Code, Azure MCP Server automates setup for vaulted backups and ransomware-protected recovery, ensuring data and cyber resiliency. A subsequent section reveals AI-powered prioritization in Azure Advisor, where recommendations are ranked based on factors such as blast radius, deadlines, and cost (<a class=\"medius-aisummary\" data-starttime=\"00:33:23\">00:33:23</a>). The Advisor Experience generates a concise, actionable plan—from hundreds of recommendations down to the top five urgent fixes—helping teams optimize remediation through intelligent ranking and integration with existing workflows like Jira or ServiceNow (<a class=\"medius-aisummary\" data-starttime=\"00:41:00\">00:41:00</a>).</p>\n\n<p><strong>Staying Resilient and Final Announcements:</strong> The closing section emphasizes the importance of continuous validation through simulated fault conditions. At Build, Microsoft announces the public preview of Azure Chaos Studio Workspaces launching June 11 (<a class=\"medius-aisummary\" data-starttime=\"00:44:00\">00:44:00</a>), a new application-centric environment for resilience testing. This tool allows customers to safely rehearse outages—such as zone failures or cascading slowdowns—to ensure recovery workflows perform as expected. The presenters conclude by reiterating that resiliency is both an engineering discipline and a sustained practice, with the new agentic tools empowering organizations to design, validate, and evolve their resilience postures directly within their daily development and operational workflows. The session ends with an invitation for audience participation in the limited preview program and a live Q&A (<a class=\"medius-aisummary\" data-starttime=\"00:44:24\">00:44:24</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/35984817-c07b-4b72-a5f0-2f8a6803325d","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/35984817-c07b-4b72-a5f0-2f8a6803325d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/35984817-c07b-4b72-a5f0-2f8a6803325d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/35984817-c07b-4b72-a5f0-2f8a6803325d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/35984817-c07b-4b72-a5f0-2f8a6803325d","speakerIds":["1768245294609build26-1775784663200001lhkW","1768245294609build26-1777047134243001sFy8","1768245294609build26-1751119338056001FUds"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/35984817-c07b-4b72-a5f0-2f8a6803325d","ogTitle":"Modern resiliency from build to recovery through Agentic AI","ogDescription":"Resiliency failures rarely come from missing features; they stem from fragmentation. During incidents teams lose time jumping between repos, dashboards, runbooks, and chat tools. This session reframes resiliency as an agent-first practice spanning build, operate, troubleshoot, and recovery. See how an agentic AI assistant links workflows across IDEs, source repos, and collaboration tools speeding validation, operations, and recovery for a critical workload aligned with Azure’s agentic model.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775784663200001lhkW","RegistrantKey":"1768245294609build26-1775784663200001lhkW"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751119338056001FUds","RegistrantKey":"1768245294609build26-1751119338056001FUds"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777047134243001sFy8","RegistrantKey":"1768245294609build26-1777047134243001sFy8"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK228","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Azure Copilot","logicalValue":"Azure Copilot"},{"displayValue":"OSS","logicalValue":"OSS"}]},{"@search.score":1.0,"documentId":"0614dda9-834d-482d-883e-e2a8888bae4a","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pierce Boggan, Kent C. Dodds, Burke Holland, Julia Kasper, Harald Kirschner, Chris Reddington","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708407001rAVo","langLocale":"en-US","sessionId":"1774018708407001rAVo","sessionInstanceId":"1774018708407001rAVo","sessionCode":"BRK201","sessionCodeNormalized":"EN-US-BRK201","title":"Multi-agent patterns in VS Code you won't learn from docs","sortTitle":"multi-agent patterns in vs code you won't learn from docs","description":"Building with one agent is familiar. Orchestrating a fleet of them in parallel across local, background, and cloud surfaces is where it gets real. You'll see the decisions that matter, how to decompose work across agents, when to fork vs. delegate, and how to verify quality when agents outnumber you. Leave with patterns for multi-agent workflows you can apply to your own codebase tomorrow.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Setup:</strong> The session opens with greetings and energy building activities as the hosts welcome the audience <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. The event is presented as a live coding competition focused on exploring multi-agent patterns within Visual Studio Code <a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>. Co-host Kent C. Dodds introduces himself as a JavaScript educator emphasizing product thinking and how AI has changed the development landscape compared to the previous year's competition, which featured simple projects like Frogger <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>. The hosts underscore AI’s rapid evolution, setting the stage for this year’s real-world challenge — building a cloud-based collaborative markdown editor <a class=\"medius-aisummary\" data-starttime=\"00:03:46\">00:03:46</a>.</p>\n\n<p><strong>Contestants and Tools Overview:</strong> The competition’s participants are introduced along with the tools they will use <a class=\"medius-aisummary\" data-starttime=\"00:03:52\">00:03:52</a>. Julia from the VS Code team works with the local agent feature in VS Code <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>, Chris from the UK codes via the Copilot CLI <a class=\"medius-aisummary\" data-starttime=\"00:04:08\">00:04:08</a>, Pierce employs the new GitHub Copilot app, and Harold tests the new agent app for multi-agent functions. The hosts clarify that each contestant’s environment represents a different facet of agent development. Meanwhile, Kent and the main host explain they will build a voting web app in parallel to allow the audience to vote live for winners <a class=\"medius-aisummary\" data-starttime=\"00:05:51\">00:05:51</a>.</p>\n\n<p><strong>Building the Voting Application:</strong> The hosts begin the technical demo by launching a GitHub Codespace environment to create their voting application safely <a class=\"medius-aisummary\" data-starttime=\"00:05:52\">00:05:52</a>. They discuss how agents operate effectively when given full permissions (“Yolo mode”) but must be sandboxed securely within cloud environments for safety <a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a>. The duo brainstorm the product requirements — a one-page site where users can vote, see results update in real time, and display a winner <a class=\"medius-aisummary\" data-starttime=\"00:07:25\">00:07:25</a>. Kent guides the design toward simplicity with vanilla JavaScript, HTML, and CSS for speed. They experiment with adding QR code functionality for mobile participation and handle library-loading issues while refining the code to create a working QR access point for real-time voting <a class=\"medius-aisummary\" data-starttime=\"00:13:14\">00:13:14</a>.</p>\n\n<p><strong>Competition Progress and Demonstrations:</strong> The scene shifts back to the contestants as the hosts check on each person’s progress. Julia implements GPT 5.5 in VS Code to scaffold her markdown application, emphasizing high reasoning and front-end improvements <a class=\"medius-aisummary\" data-starttime=\"00:16:05\">00:16:05</a>. Chris uses an orchestrated multi-agent approach with GPT and Codecs to divide planning and implementation tasks for real-time collaboration <a class=\"medius-aisummary\" data-starttime=\"00:17:19\">00:17:19</a>. Pierce performs experimentation inside the Copilot app with markdown rendering, emojis, and AI-driven text generation <a class=\"medius-aisummary\" data-starttime=\"00:31:40\">00:31:40</a>. Harold showcases a multi-agent approach involving research agents and design exploration, producing prototypes that simulate collaborative editing with commenting and version control features <a class=\"medius-aisummary\" data-starttime=\"00:20:39\">00:20:39</a>.</p>\n\n<p><strong>Voting Integration and Advancing Features:</strong> Kent and the host successfully debug and deploy the voting app for public access via QR code <a class=\"medius-aisummary\" data-starttime=\"00:21:19\">00:21:19</a>. They discuss persistence mechanisms using JSON files, event-driven updates, and the eventual integration of WebSockets to synchronize votes across users <a class=\"medius-aisummary\" data-starttime=\"00:25:23\">00:25:23</a>. Additional design iterations are executed rapidly using multiple agents to generate UI mock-ups, informed by the host’s custom “poster board aesthetic” design system <a class=\"medius-aisummary\" data-starttime=\"00:35:20\">00:35:20</a>. They compare approaches to parallelizing design generation and show iterative styling output within their agent-driven workspace <a class=\"medius-aisummary\" data-starttime=\"00:36:13\">00:36:13</a>.</p>\n\n<p><strong>Final Presentations and Conclusion:</strong> As time winds down, each contestant presents their final work. Julia’s split-screen markdown editor features live collaboration, username customization, and emoji-based vibes <a class=\"medius-aisummary\" data-starttime=\"00:41:47\">00:41:47</a>. Chris’s version introduces a sidebar, multi-file management, and GitHub-themed styling through Primer library components <a class=\"medius-aisummary\" data-starttime=\"00:42:22\">00:42:22</a>. Pierce adds playful user interaction via emoji reactions and comments <a class=\"medius-aisummary\" data-starttime=\"00:43:29\">00:43:29</a>, while Harold integrates advanced agent-based prototypes combining design, architecture analysis, and planning histories into a sophisticated collaborative environment <a class=\"medius-aisummary\" data-starttime=\"00:44:21\">00:44:21</a>. When voting concludes via the live app, Harold emerges as the winner <a class=\"medius-aisummary\" data-starttime=\"00:45:59\">00:45:59</a>. The closing remarks emphasize the difficulty of live coding, appreciation for the participants, and encourage attendees to join upcoming Copilot and multi-agent sessions, thanking everyone for attending <a class=\"medius-aisummary\" data-starttime=\"00:46:13\">00:46:13</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"18:30 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/c3a80714-e18b-4cd1-b75c-92204129a14d","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c3a80714-e18b-4cd1-b75c-92204129a14d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c3a80714-e18b-4cd1-b75c-92204129a14d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c3a80714-e18b-4cd1-b75c-92204129a14d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c3a80714-e18b-4cd1-b75c-92204129a14d","speakerIds":["1768245294609build26-1775704791363001hLZq","1768245294609build26-1779392564366001On0T","1768245294609build26-1776347378057001Rvny","1768245294609build26-1774303544298001acRa","1768245294609build26-1774285000231001pWGh","1768245294609build26-1775629342327001E2I8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.465Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c3a80714-e18b-4cd1-b75c-92204129a14d","ogTitle":"Multi-agent patterns in VS Code you won't learn from docs","ogDescription":"Building with one agent is familiar. Orchestrating a fleet of them in parallel across local, background, and cloud surfaces is where it gets real. You'll see the decisions that matter, how to decompose work across agents, when to fork vs. delegate, and how to verify quality when agents outnumber you. Leave with patterns for multi-agent workflows you can apply to your own codebase tomorrow.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774303544298001acRa","RegistrantKey":"1768245294609build26-1774303544298001acRa"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000231001pWGh","RegistrantKey":"1768245294609build26-1774285000231001pWGh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347378057001Rvny","RegistrantKey":"1768245294609build26-1776347378057001Rvny"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779392564366001On0T","RegistrantKey":"1768245294609build26-1779392564366001On0T"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775629342327001E2I8","RegistrantKey":"1768245294609build26-1775629342327001E2I8"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775704791363001hLZq","RegistrantKey":"1768245294609build26-1775704791363001hLZq"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK201","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"b12012cb-def8-47f4-bcf0-2b694574d5ea","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sarah Bird, Sandeep Atluri, Mehrnoosh Sameki, Katelyn Rothney","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018716553001rOhY","langLocale":"en-US","sessionId":"1774018716553001rOhY","sessionInstanceId":"1774018716553001rOhY","sessionCode":"BRK250","sessionCodeNormalized":"EN-US-BRK250","title":"Observe and control agents across any framework with open source tools","sortTitle":"observe and control agents across any framework with open source tools","description":"As AI agents move into production, developers own safety, governance, and reliability across Microsoft Agent Framework and open-source stacks. This session shows how to govern agents end to end: turning your requirements into context-aware evaluations, stress-testing against adversarial risks, applying open controls that work across frameworks, and keeping humans in the loop on high-stakes actions. Leave with a blueprint for shipping agents at enterprise scale.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Responsible AI Context:</strong> The video opens with Sarah Byrd introducing herself as Microsoft's Chief Product Officer for Responsible AI, alongside Sandeep, who leads Responsible AI science efforts <a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>. Byrd begins by highlighting the increasing risks of autonomous agents in practical systems where errors can create company-wide and societal impact <a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>. She cites research showing 60% of agents have privileged data access, often causing unauthorized sharing and security breaches. Byrd emphasizes Microsoft’s active development of tools to observe, control, and secure AI agents across development and production phases.</p>\n\n<p><strong>Understanding Agent Failures and the “Lethal Trifecta”:</strong> Byrd explains four major categories of agent failure: misunderstanding instructions, hallucinating information, leaking sensitive data, and interacting unpredictably in multi-agent systems <a class=\"medius-aisummary\" data-starttime=\"00:01:56\">00:01:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:03\">00:03:03</a>. She expands on a common pattern termed the “lethal trifecta,” wherein context rot, confusion, and privilege misuse can lead agents to exfiltrate confidential information <a class=\"medius-aisummary\" data-starttime=\"00:03:18\">00:03:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:25\">00:04:25</a>. To mitigate these, she outlines the foundational Responsible AI process loop—identify risks, evaluate them, apply controls, and continuously refine based on performance outcomes <a class=\"medius-aisummary\" data-starttime=\"00:04:43\">00:04:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:51\">00:05:51</a>.</p>\n\n<p><strong>Demo: Banking Agent Risk and Evaluation with ASSERT:</strong> Sandeep introduces a demo featuring a “Bank Manager Agent” designed for account balances and money transfers, built using LangGraph and MCP Server <a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:06\">00:07:06</a>. Byrd and Sandeep discuss domain-specific risks including prompt injection, hallucinations, and unauthorized transfers <a class=\"medius-aisummary\" data-starttime=\"00:07:17\">00:07:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>. Byrd then announces Microsoft’s open-source tool ASSERT (Agent Systematic Evaluation and Risk Testing), enabling organizations to define evaluation metrics aligned with their safety requirements <a class=\"medius-aisummary\" data-starttime=\"00:10:22\">00:10:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:08\">00:11:08</a>. ASSERT converts broad Responsible AI principles into granular testable rules and automatically generates falsifiable evaluations for agent behaviors.</p>\n\n<p><strong>ASSERT Demo Results and Integration:</strong> In the live demo, Sandeep shows how developers use ASSERT by writing simple YAML definitions of risks like “do not leak sensitive data or execute unauthorized transactions” <a class=\"medius-aisummary\" data-starttime=\"00:13:18\">00:13:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:02\">00:14:02</a>. The framework systematizes these into categories and test sets covering both single-turn and multi-turn conversations <a class=\"medius-aisummary\" data-starttime=\"00:15:02\">00:15:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:21\">00:17:21</a>. The initial evaluation shows baseline policy violations of 28–58%, later reduced with improved prompts. Though prompting offers limited effectiveness in layered scenarios, integrating deterministic guardrails through ACS—Agent Control Specification—drives a reduction toward near-zero violation rates <a class=\"medius-aisummary\" data-starttime=\"00:27:14\">00:27:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:30:07\">00:30:07</a>. The conversation underscores the balance between avoiding policy violations and preventing over-refusal issues to preserve customer experience.</p>\n\n<p><strong>ACS and Production Governance with Foundry:</strong> Byrd introduces ACS, an open-source Agent Control Specification designed to unify control logic and provide deterministic as well as AI-assisted safety checks across frameworks <a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>. Integrated into the Agent Governance Toolkit (AGT) and Microsoft Foundry, ACS ensures consistent enforcement of input/output, data governance, and tool-specific rules in production settings. Foundry ties ACS with Microsoft Defender and Purview for data protection and integrates identity and tracing features for full observability <a class=\"medius-aisummary\" data-starttime=\"00:34:00\">00:34:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:35:54\">00:35:54</a>. Byrd emphasizes continuous monitoring using ASSERT and ACS in live traffic for iterative improvement and alignment with organizational policies.</p>\n\n<p><strong>Future Frontiers and Responsible Collaboration:</strong> In closing, Byrd and Sandeep preview upcoming advancements including continuous evaluations powered by reinforcement-learning attackers that adapt tests based on production performance <a class=\"medius-aisummary\" data-starttime=\"00:37:03\">00:37:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:39:20\">00:39:20</a>. They highlight work on content authenticity via invisible C2PA watermarks <a class=\"medius-aisummary\" data-starttime=\"00:41:11\">00:41:11</a> and deterministic data flow policy enforcement <a class=\"medius-aisummary\" data-starttime=\"00:42:20\">00:42:20</a>. The session ends with a call for open collaboration across the community to improve AI trust, emphasizing transparent evaluation, control, and continuous innovation <a class=\"medius-aisummary\" data-starttime=\"00:45:09\">00:45:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:46\">00:45:46</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:30 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","speakerIds":["1768245294609build26-1775085931036001BzcI","1768245294609build26-1776099704878001Zi6e","1768245294609build26-1775085874620001ADSc","1768245294609build26-1756838348665001FzIo"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.855Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b919f3dd-95e5-41c4-9ea5-7604e8d87df8","ogTitle":"Observe and control agents across any framework with open source tools","ogDescription":"As AI agents move into production, developers own safety, governance, and reliability across Microsoft Agent Framework and open-source stacks. This session shows how to govern agents end to end: turning your requirements into context-aware evaluations, stress-testing against adversarial risks, applying open controls that work across frameworks, and keeping humans in the loop on high-stakes actions. Leave with a blueprint for shipping agents at enterprise scale.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1756838348665001FzIo","RegistrantKey":"1768245294609build26-1756838348665001FzIo"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775085931036001BzcI","RegistrantKey":"1768245294609build26-1775085931036001BzcI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776099704878001Zi6e","RegistrantKey":"1768245294609build26-1776099704878001Zi6e"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775085874620001ADSc","RegistrantKey":"1768245294609build26-1775085874620001ADSc"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}]},{"@search.score":1.0,"documentId":"12d25e75-33a7-4d96-bc6a-d07174be26c5","startDateTime":"2026-06-02T22:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Joey Conway, AYSEN Ilkbahar, Stephen McCullough","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775842562865001MGWT","langLocale":"en-US","sessionId":"1775842562865001MGWT","sessionInstanceId":"1775842562865001MGWT","sessionCode":"BRKSP94","sessionCodeNormalized":"EN-US-BRKSP94","title":"Orchestrate special agents with NVIDIA Nemotron models on Foundry","sortTitle":"orchestrate special agents with nvidia nemotron models on foundry","description":"Optimize enterprise Agentic AI with a tiered system-of-models architecture in Microsoft Foundry. This session demonstrates a plan-and-execute pattern routing tasks across frontier models for reasoning, NVIDIA Nemotron for complex sub-tasks, and local models for latency-sensitive execution. Learn to route workloads across cloud and edge tiers to minimize cost-per-task while maximizing quality. Dive into using special agents and post-trained open-source models to achieve faster task completion.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The video opens with Joey Conway confirming audio and stating that the session will cover new announcements from NVIDIA, collaborations with partners such as Microsoft, and ongoing work with agent orchestration (<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>). He shares the agenda: a brief history of AI model development, details on NVIDIA’s Nemotron open model family, and the company’s efforts to strengthen the open ecosystem (<a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>). Conway then recalls the early AI milestones, starting from the ChatGPT moment four years ago, followed by the release of DeepSeek reasoning models, and finally the emergence of agentic systems, which enable models to interact through complex task execution rather than simple Q&A (<a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>).</p>\n\n<p><strong>Understanding Agent Systems:</strong> Conway explains NVIDIA’s conceptual framework for long-running agentic systems, describing a continuous loop where agents receive prompts, analyze context, observe data, create plans, act, and iterate until tasks are complete (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>). He details the components involved, such as tools, skills, memory, and orchestration between multiple agents. These systems are powered by large language models (LLMs) and leverage enterprise-level security, governance, and CUDA-based computation across CPU and GPU platforms. Conway emphasizes NVIDIA’s vision of agentic systems of models—multiple specialized AI models working together—with an emphasis on efficiency across pre-training, fine-tuning, and inference phases, always striving for reduced compute costs without sacrificing intelligence (<a class=\"medius-aisummary\" data-starttime=\"00:04:26\">00:04:26</a>).</p>\n\n<p><strong>Nemotron Family Overview:</strong> Conway introduces the NVIDIA Nemotron model family, describing it as an open suite of models, tools, and datasets for reasoning and multimodal intelligence (<a class=\"medius-aisummary\" data-starttime=\"00:05:47\">00:05:47</a>). He outlines five main pillars: reasoning, vision, information retrieval, safety moderation, and speech processing (<a class=\"medius-aisummary\" data-starttime=\"00:06:01\">00:06:01</a>). Conway transitions into the specifics of Nemotron’s three scalable tiers—Nano, Super, and Ultra—each optimized for different data center footprints and performance needs (<a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>). Nano targets compact GPU setups, Super is designed for mid-scale enterprise tasks, and Ultra represents NVIDIA’s most capable model, optimized for frontier reasoning, orchestration tasks, and reduced inference cost. Announced recently, Nemotron 3 Ultra integrates advanced training techniques like multi-teacher distillation (<a class=\"medius-aisummary\" data-starttime=\"00:09:08\">00:09:08</a>) and new NVFP4 precision formats for NVIDIA Blackwell architecture, allowing one unified checkpoint for device compatibility.</p>\n\n<p><strong>Nemotron 3 Ultra Capabilities and Performance:</strong> Conway presents Nemotron 3 Ultra as NVIDIA’s most advanced open model built with hybrid Mixture-of-Experts (MoE) architecture, multi-token prediction, and one-million-token context length (<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>). He explains how it achieves faster inference, improved accuracy, and cost-efficient reasoning through model distillation and hardware-optimized compute. Benchmark analyses demonstrate Ultra’s dominance in agentic tasks such as complex coding, long-horizon enterprise operations, and professional-level reasoning (<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>). Third-party evaluations by Artificial Analysis confirm the model delivers frontier accuracy alongside fivefold throughput improvements. Conway concludes this segment emphasizing that Ultra moves models toward the cost-efficient frontier—high accuracy at low compute cost—representing optimal task and resource tradeoffs (<a class=\"medius-aisummary\" data-starttime=\"00:15:02\">00:15:02</a>).</p>\n\n<p><strong>Demo of Hermes Agent and Enterprise Workflow:</strong> The session transitions into a Nemotron 3 Ultra demo illustrating AI orchestration in real business scenarios with Hermes Agent (<a class=\"medius-aisummary\" data-starttime=\"00:16:02\">00:16:02</a>). Conway describes how a small company’s team collaborates with Hermes, which learns and refines reusable skills, creating organizational \"collective wisdom\" as these skills propagate to other users. He explains that over time, employee-generated skills accelerate workflow and enable smarter digital co-workers. Conway concludes his segment predicting a future where employees evaluate companies by the quality of their digital agents and productivity support (<a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>).</p>\n\n<p><strong>Microsoft Foundry Integration and Closing:</strong> Stephen McCullough takes over and explores NVIDIA’s partnership with Microsoft to deploy long-running agents through Microsoft Foundry-hosted environments (<a class=\"medius-aisummary\" data-starttime=\"00:19:10\">00:19:10</a>). He demonstrates how hosted Hermes Agents integrate Nemotron models with enterprise resources like Outlook, GitHub, and Teams, using Entra ID identities to manage access securely. Through a debugged scenario, McCullough shows Hermes receiving an email, creating a GitHub PR, applying feedback, and learning new coding skills that persist across sessions and platforms, all visible through Foundry’s observability tools (<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>). He concludes that Foundry enables enterprises to treat agents like human employees—granting, auditing, and revoking permissions—while continuously improving workflows. McCullough ends by summarizing the combined value of Nemotron’s intelligence, Hermes’ orchestration, and Foundry’s enterprise-grade governance, inviting viewers to visit NVIDIA’s booth, explore their open resources, and participate in the feedback survey (<a class=\"medius-aisummary\" data-starttime=\"00:37:17\">00:37:17</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"22:45 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/6c7fb279-6292-41b9-8bf5-5d5b12d9eb25","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/6c7fb279-6292-41b9-8bf5-5d5b12d9eb25","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/6c7fb279-6292-41b9-8bf5-5d5b12d9eb25","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/6c7fb279-6292-41b9-8bf5-5d5b12d9eb25","speakerIds":["1768245294609build26-1778610401464001RasM","1768245294609build26-1778172041769001TfQN","1768245294609build26-1769795191086001hVae"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.901Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/6c7fb279-6292-41b9-8bf5-5d5b12d9eb25","ogTitle":"Orchestrate special agents with NVIDIA Nemotron models on Foundry","ogDescription":"Optimize enterprise Agentic AI with a tiered system-of-models architecture in Microsoft Foundry. This session demonstrates a plan-and-execute pattern routing tasks across frontier models for reasoning, NVIDIA Nemotron for complex sub-tasks, and local models for latency-sensitive execution. Learn to route workloads across cloud and edge tiers to minimize cost-per-task while maximizing quality. Dive into using special agents and post-trained open-source models to achieve faster task completion.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778172041769001TfQN","RegistrantKey":"1768245294609build26-1778172041769001TfQN"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769795191086001hVae","RegistrantKey":"1768245294609build26-1769795191086001hVae"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778610401464001RasM","RegistrantKey":"1768245294609build26-1778610401464001RasM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"d9c9d15a-e94d-4c94-8b87-aadc2eb79ac3","startDateTime":"2026-06-03T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Krunal Patel, Rishabh Saha, Bob Ward, Govanna Flores","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018704684001rpfv","langLocale":"en-US","sessionId":"1774018704684001rpfv","sessionInstanceId":"1774018704684001rpfv","sessionCode":"BRK224","sessionCodeNormalized":"EN-US-BRK224","title":"PepsiCo’s blueprint for agentic AI","sortTitle":"pepsico’s blueprint for agentic ai","description":"Learn from Microsoft and PepsiCo engineers who modernized PepsiCo’s data foundation for agentic applications using Azure SQL, Cosmos DB, PostgreSQL, and Azure Databricks. Discover a practical build path for agentic RAG architecture, leveraging Azure SQL features like vector indexing and semantic search, to streamline app development and enable faster, repeatable patterns.  Refresh your own data layer to reduce app development cycles while leveraging modern, repeatable patterns to ship faster.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens at <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a> with Bob Ward welcoming attendees to Microsoft Build and reflecting on the array of new technologies showcased, especially around AI and data. He emphasizes the real value in seeing customers succeed with technological solutions. Bob introduces himself as a Principal Architect on the Azure Data team and outlines the focus of the presentation — a collaboration between Microsoft and PepsiCo to develop an AI-driven data solution. Before bringing PepsiCo representatives on stage, Bob discusses the foundational need for proper data infrastructure in AI projects, referencing Gartner studies at <a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a> that show AI initiatives often fail without high-quality, secure data. He illustrates the common “starting line” scenario of fragmented, siloed data and vendor sprawl, explaining how unified lakehouse architectures powered by Microsoft Fabric and Azure Databricks can lead organizations toward “data nirvana” with integrated governance, security, and AI readiness.</p>\n\n<p><strong>Data Modernization and AI Readiness:</strong> Continuing at <a class=\"medius-aisummary\" data-starttime=\"00:02:49\">00:02:49</a>, Bob describes the journey companies must take to unify their data. He explains that platforms like Microsoft Fabric and Databricks simplify the process of integrating data and making it AI-ready, emphasizing that curated, secure, high-quality data is essential. He outlines stages of extending AI use — from unified data to fine-tuned models to various deployment scenarios, including edge solutions using Foundry Local. As the presentation transitions, Bob showcases Azure database offerings at <a class=\"medius-aisummary\" data-starttime=\"00:05:17\">00:05:17</a>, describing SQL Hyperscale, PostgreSQL, Cosmos DB, and Horizon DB, highlighting their scalability, transactional integrity, and integrated vector search capabilities for AI. These tools together lay the foundation for PepsiCo’s innovations in data and AI integration before bringing the PepsiCo case study to the stage.</p>\n\n<p><strong>PepsiCo’s Scale and Foundations for AI Transformation:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>, Rishab Saha from Microsoft and Krunal Patel from PepsiCo begin their presentation, starting with PepsiCo’s immense operational scale and the complexity that accompanies global data. They introduce the central challenge faced by key account managers (Cams): fragmented information across numerous systems and reports, described vividly between <a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>. Krunal imagines a world with unified, intelligent workflows and learning systems that evolve without continuous manual data input. The discussion moves to PepsiCo’s Enterprise Data Foundation (EDF), a robust governed data layer ensuring consistency, harmonization, and standardization across markets. Rishab explains EDF’s critical role in transforming data for dashboards into data for AI agents, supporting connected knowledge graphs and intelligent agents capable of reasoning and acting autonomously.</p>\n\n<p><strong>Agent Architecture and Demonstrations:</strong> The session deepens between <a class=\"medius-aisummary\" data-starttime=\"00:17:05\">00:17:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:04\">00:25:04</a> with examples of PepsiCo’s Cam 360 agent system. Using a persona named Priya, the presenters illustrate how multiple agents — orchestrator, analyst, tactical, lifecycle tracking, best practices, and conversational debrief agents — assist a human account manager. Each agent has distinct roles, from crunching data to maintaining account lifecycles. The data analyst (NERD) agent demo showcases a text-to-SQL translation pipeline built on Databricks and Genie, using Unity Catalog for access control and high-confidence query paths derived from real-user interactions. Krunal explains authentication layers, SQL generation, golden datasets, and observability mechanisms ensuring governance and accuracy. This section highlights how PepsiCo combined Microsoft technologies — LangSmith, Fabric, and Databricks — to create secure, automated data workflows that respond intelligently to natural language queries.</p>\n\n<p><strong>From Preparation to Learning: Tracking and Compound Intelligence:</strong> Between <a class=\"medius-aisummary\" data-starttime=\"00:27:32\">00:27:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:00\">00:36:00</a>, Rishab introduces the tracking agent and explains how it transforms individual knowledge into institutional memory using Foundry IQ and Cosmos DB for learning signals and configurable categories. The fact builder agent consolidates these signals and knowledge bases into reliable facts, generating summaries with confidence scores and contextual data. The tracking agent interface allows users to instantly review prior interactions or customer histories through conversational queries while telemetry ensures real-time observability. This system embodies “compounding intelligence”—each meeting and debrief builds on previous insights, improving future decisions and capturing tribal knowledge into structured enterprise intelligence. Priya’s workflow now evolves from manual preparation to intelligent, automated understanding, saving time and increasing strategic focus.</p>\n\n<p><strong>Key Lessons and Blueprint for Enterprise AI:</strong> The final segment at <a class=\"medius-aisummary\" data-starttime=\"00:37:03\">00:37:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:27\">00:41:27</a> reviews lessons learned. Rishab and Krunal advise starting small, ensuring domain experts help validate data quality, and investing deeply in foundational data architecture before scaling agents. They reflect on how once initial agents were built, subsequent ones were far quicker due to solid foundations. They conclude with a blueprint for enterprise AI — secure infrastructure, unified and governed data, AI platforms ready with chosen models, and seamless user interaction layers. The PepsiCo example becomes a scalable model for any enterprise undergoing AI transformation. Closing at <a class=\"medius-aisummary\" data-starttime=\"00:41:15\">00:41:15</a>, they revisit Priya’s workflow: preparing for customers now takes minutes instead of hours, freeing teams to focus on strategy and relationships. The presenters affirm that AI’s role is not to replace people but to give time back so they can think, innovate, and be human.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:45 - 22:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/95afb175-5c57-46c2-bd53-d67dda3d0888","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/95afb175-5c57-46c2-bd53-d67dda3d0888","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/95afb175-5c57-46c2-bd53-d67dda3d0888","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/95afb175-5c57-46c2-bd53-d67dda3d0888","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/95afb175-5c57-46c2-bd53-d67dda3d0888","speakerIds":["1768245294609build26-1778785147201001uJOv","1768245294609build26-1776447112591001abTN","1768245294609build26-1765362751973001nhSz","1768245294609build26-1771864681786001rYr8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.459Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/95afb175-5c57-46c2-bd53-d67dda3d0888","ogTitle":"PepsiCo’s blueprint for agentic AI","ogDescription":"Learn from Microsoft and PepsiCo engineers who modernized PepsiCo’s data foundation for agentic applications using Azure SQL, Cosmos DB, PostgreSQL, and Azure Databricks. Discover a practical build path for agentic RAG architecture, leveraging Azure SQL features like vector indexing and semantic search, to streamline app development and enable faster, repeatable patterns.  Refresh your own data layer to reduce app development cycles while leveraging modern, repeatable patterns to ship faster.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1771864681786001rYr8","RegistrantKey":"1768245294609build26-1771864681786001rYr8"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362751973001nhSz","RegistrantKey":"1768245294609build26-1765362751973001nhSz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776447112591001abTN","RegistrantKey":"1768245294609build26-1776447112591001abTN"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778785147201001uJOv","RegistrantKey":"1768245294609build26-1778785147201001uJOv"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK224","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"14b159af-7446-464d-8566-a096b7e1ce79","startDateTime":"2026-06-03T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vijay Aski, Manoj Bableshwar, Chris Lauren","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018713033001raz6","langLocale":"en-US","sessionId":"1774018713033001raz6","sessionInstanceId":"1774018713033001raz6","sessionCode":"BRK232","sessionCodeNormalized":"EN-US-BRK232","title":"Post-Training and Deploying Open Source Reasoning Models in Foundry","sortTitle":"post-training and deploying open source reasoning models in foundry","description":"Open-source reasoning models are powerful out of the box, but production performance comes from closing the loop. Learn how to use Microsoft Foundry to collect production traces, curate them into datasets, and post-train reasoning models with RL using frameworks like slime, verl, and TRL. We'll cover when RL drives real gains and how to redeploy improved models without managing infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The session opens with Chris Lauren welcoming attendees and outlining the objective of converting them into experts in post-training open-source models and deploying AI agents at scale <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>. Chris highlights that deploying a model into production is only the beginning and stresses the importance of continuously learning, improving, and fine-tuning models to perform well across scenarios. The goal is to achieve faster, better, and more cost-efficient agents using Microsoft Foundry. He explains the growing importance of agents in enterprise AI transformation and introduces the challenges of token consumption and costs when scaling AI workloads <a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:10\">00:02:10</a>.</p>\n\n<p><strong>Fine-Tuning, Evaluation, and Model Selection:</strong> Chris transitions to discussing post-training or fine-tuning as crucial to customizing models for specific business needs <a class=\"medius-aisummary\" data-starttime=\"00:02:28\">00:02:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:20\">00:04:20</a>. He elaborates that fine-tuning teaches models appropriate tool use, adherence to company policies, and domain understanding, ultimately reducing costly missteps. Foundry provides a platform with a wide model catalog and built-in evaluation mechanisms. Through evaluation, users can determine the most cost-effective and high-quality model for their particular agent. To illustrate, Chris demonstrates a retail customer-service agent scenario that compares frontier models such as GPT-5 with open-source models like Qwen3-14B, emphasizing Foundry’s ability to trace performance, manage cost, and refine models by iterating on evaluations <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>.</p>\n\n<p><strong>Evaluators and Continuous Improvement with Foundry:</strong> The conversation deepens into the concept of evaluators, which Chris defines as measurable specifications that guide what “good performance” looks like <a class=\"medius-aisummary\" data-starttime=\"00:09:10\">00:09:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>. He explains that benchmarks from external sources do not capture enterprise-specific definitions of success. Foundry allows both built-in and custom evaluations that map directly to a company’s policies and quality criteria. Chris demonstrates evaluating multiple agents and identifying quality improvement paths through iterative refinement. He highlights how Foundry enables users to “hill climb” by testing, measuring, and optimizing agents based on custom business metrics, leading to tangible quality and cost gains before turning the stage over to Vijay Aski <a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:40\">00:12:40</a>.</p>\n\n<p><strong>Post-Training and Reinforcement Learning Techniques:</strong> Vijay introduces post-training as an umbrella term that includes fine-tuning and reinforcement learning from human feedback (RLHF) <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>. He explains three model training stages—pre-training, mid-training, and post-training—emphasizing that post-training optimizes a model for specific, verifiable agentic tasks. Vijay demonstrates how Foundry helps teams use production traces, human preferences, and synthetic data for distillation and structured learning. He outlines two main workflows: supervised fine-tuning (SFT), where models mimic experts or ideal behaviors, and reinforcement fine-tuning (RFT), which rewards correct tool usage and accurate task completion. Foundry’s unified system lets data scientists monitor GPU clusters, training jobs, and real-time rollouts with full visibility, enabling systematic optimization and consistent quality improvement <a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:30:00\">00:30:00</a>.</p>\n\n<p><strong>Advanced Training, APIs, and Simplified Compute Management:</strong> Vijay continues by presenting Foundry’s dual offerings for advanced users and non-specialists <a class=\"medius-aisummary\" data-starttime=\"00:30:00\">00:30:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:00\">00:36:00</a>. Code-first experiences in Foundry allow for deep configuration of Ray clusters, GPUs, and model training parameters, while upcoming low-level APIs—code-named “Loom”—simplify training with minimal infrastructure setup. The APIs abstract away cluster management while exposing essential GPU primitives for both supervised and reinforcement fine-tuning. Both interfaces provide rich monitoring dashboards with loss, reward, and entropy metrics to observe learning progress. Vijay concludes by showcasing Foundry’s seamless integration between traditional and API-first workflows, enabling iterative training loops and adaptive infrastructure scaling for different use cases.</p>\n\n<p><strong>Deployment and Conclusion:</strong> Chris returns to introduce the deployment component and announces Foundry Managed Compute—a scalable serverless offering for deploying any custom or open-source model <a class=\"medius-aisummary\" data-starttime=\"00:37:00\">00:37:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:46:40\">00:46:40</a>. Manoj Bableshwar demonstrates various deployment options, from base to fine-tuned models, using Managed Compute’s GPU configurations and integration into agents for live retrieval, tool calling, and contextual memory tasks. He also presents dashboards for real-time monitoring, observability, and cost-tracking. Chris concludes by summarizing the key takeaways: Foundry’s unified workflow supports model evaluation, fine-tuning, post-training, and deployment at scale with full telemetry and cost management. Attendees are encouraged to experiment immediately using published repositories and Foundry’s tools to continuously improve models and agents in production environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:45 - 22:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:30:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","speakerIds":["1768245294609build26-1776895269615001aQWT","1768245294609build26-1775238034507001eQ13","1768245294609build26-1776456842448001LPgI"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.853Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b7ec6cc5-da1b-46e4-bce5-748e0e07c737","ogTitle":"Post-Training and Deploying Open Source Reasoning Models in Foundry","ogDescription":"Open-source reasoning models are powerful out of the box, but production performance comes from closing the loop. Learn how to use Microsoft Foundry to collect production traces, curate them into datasets, and post-train reasoning models with RL using frameworks like slime, verl, and TRL. We'll cover when RL drives real gains and how to redeploy improved models without managing infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776456842448001LPgI","RegistrantKey":"1768245294609build26-1776456842448001LPgI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775238034507001eQ13","RegistrantKey":"1768245294609build26-1775238034507001eQ13"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776895269615001aQWT","RegistrantKey":"1768245294609build26-1776895269615001aQWT"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK232","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e50162cd-c9eb-4295-8015-e7a13414e46e","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Colin Helms, Imran Sheik Mohamed, Karthik Vijayan, Jayneel Vora","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774983447536001GeJt","langLocale":"en-US","sessionId":"1774983447536001GeJt","sessionInstanceId":"1774983447536001GeJt","sessionCode":"BRKSP92","sessionCodeNormalized":"EN-US-BRKSP92","title":"Scale agentic AI from on-device to cloud orchestration","sortTitle":"scale agentic ai from on-device to cloud orchestration","description":"Modern AI systems span client, edge, and cloud rather than a single model. Explore how agentic AI workloads operate across these environments through three demos: real‑time on‑device agents, distributed inference across edge systems, and enterprise‑scale multi‑agent orchestration on Azure Kubernetes Service with Intel Xeon. Leave with practical guidance on placing inference, reasoning, and orchestration where they make the most sense to balance responsiveness, scale, and efficiency.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with host Eddie from Intel confirming audio and welcoming the audience of developers, engineers, and architects <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>. He introduces the theme: AI applications have grown beyond single models to distributed agentic systems that span across client, edge, and cloud infrastructures <a class=\"medius-aisummary\" data-starttime=\"00:00:21\">00:00:21</a>. The agenda includes three demos illustrating these layers — a client-side AI demo, an edge computing setup, and a cloud-scale orchestration demonstration on Azure <a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>.</p>\n\n<p><strong>Client-Side AI Demonstration:</strong> Eddie introduces J. Neal, who begins by illustrating common AI cost limitations — many developers exceed monthly budgets due to token-based usage costs <a class=\"medius-aisummary\" data-starttime=\"00:01:34\">00:01:34</a>. Neal presents Intel’s “Ion 1.0 Instant” technology released recently, running on the Panther Lake NPU of the Core Ultra Series 3 <a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>. He shows a local chat interface where responses such as identifying “the color of the sky” demonstrate the light NPU activity and minimal CPU utilization <a class=\"medius-aisummary\" data-starttime=\"00:03:13\">00:03:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>. This proof highlights near-instant inference with low power consumption and no network dependency. Neal emphasizes the quantization capability of Panther Lake — enabling compact yet powerful models without compromises on speed or intelligence <a class=\"medius-aisummary\" data-starttime=\"00:04:22\">00:04:22</a>. His demo underlines the importance of local AI execution for edge-sensitive environments like hospitals and aircraft, using OS-level APIs and OpenVINO for seamless NPU scheduling <a class=\"medius-aisummary\" data-starttime=\"00:05:29\">00:05:29</a>. He concludes inviting attendees to view the reasoning-capability game demo at Intel’s booth <a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>.</p>\n\n<p><strong>Edge Computing and Resource Pooling:</strong> Eddie next introduces Colin to explain resource pooling across AI stacks <a class=\"medius-aisummary\" data-starttime=\"00:06:49\">00:06:49</a>. Colin presents three Asus NUC Pro mini PCs powered by Intel Core Ultra Series 3 processors and Arc B390 integrated graphics <a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>. He explains how memory sharing settings from Intel’s GPU driver enable up to 93% allocation for video RAM, achieving 51 GB of VRAM per device <a class=\"medius-aisummary\" data-starttime=\"00:08:41\">00:08:41</a>. Using Llama CPP’s RPC, he connects all three devices over Thunderbolt 4 for high-speed 20 Gb/s networking, effectively pooling their resources for approximately 150 GB total VRAM <a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>. Colin demonstrates an 180-billion parameter model running locally at 16–18 tokens per second <a class=\"medius-aisummary\" data-starttime=\"00:10:45\">00:10:45</a>. He highlights cost efficiency — a fully functioning, high-performance AI stack under $7000 capable of running developer-sized workloads without network fees or cloud tokens <a class=\"medius-aisummary\" data-starttime=\"00:11:01\">00:11:01</a>. The Austin Copilot CLI integration allows pointing Visual Studio code checks to local models, showcasing full offline dev setup <a class=\"medius-aisummary\" data-starttime=\"00:12:10\">00:12:10</a>. Colin also introduces Microsoft’s MXC execution containers for safe agent sandboxing that enterprise administrators can enforce by policy <a class=\"medius-aisummary\" data-starttime=\"00:13:17\">00:13:17</a>. A script-run demo displays containerized execution behavior, highlighting secure, isolated workflows <a class=\"medius-aisummary\" data-starttime=\"00:14:21\">00:14:21</a>. His segment closes by stressing the concept of “free tokens” through local processing, enabling developers cost-free model interaction <a class=\"medius-aisummary\" data-starttime=\"00:16:45\">00:16:45</a>.</p>\n\n<p><strong>Cloud and Data Center Demonstration:</strong> The final presenter, Imran, builds upon the previous sessions by showcasing deployment at cloud scale using Azure virtual VMs <a class=\"medius-aisummary\" data-starttime=\"00:17:21\">00:17:21</a>. He introduces “Agent AI” as an evolved interaction paradigm for large language models, now capable of executing complex action sequences instead of simple chat responses <a class=\"medius-aisummary\" data-starttime=\"00:17:46\">00:17:46</a>. His demo involves hosting a local 3.635B parameter model on Azure using 96 virtual CPUs and 200 GB of RAM <a class=\"medius-aisummary\" data-starttime=\"00:19:08\">00:19:08</a>. The system utilizes Intel Xeon 6 CPUs with AMX acceleration for matrix multiplications, enabling efficient transformer inference <a class=\"medius-aisummary\" data-starttime=\"00:19:56\">00:19:56</a>. He demonstrates integration with OpenClaw running on Kubernetes, showing agents executing GitHub queries and summarizing pull requests through coordinated commands between the LLM and OpenClaw engine <a class=\"medius-aisummary\" data-starttime=\"00:22:01\">00:22:01</a>. Imran explains auto-scaling behavior — when CPU utilization increases, Kubernetes automatically scales the agent replicas from two instances upward <a class=\"medius-aisummary\" data-starttime=\"00:24:21\">00:24:21</a>. This dynamic scaling allows multiple agents to act on different personas or roles within the same system, performing simultaneous PR reviews and tasks <a class=\"medius-aisummary\" data-starttime=\"00:25:02\">00:25:02</a>. He concludes highlighting affordability and performance advantages of hosting private inference on Xeon-powered VMs <a class=\"medius-aisummary\" data-starttime=\"00:25:29\">00:25:29</a>.</p>\n\n<p><strong>Conclusion and Q&A:</strong> Eddie returns to close the presentations with a brief Q&A session <a class=\"medius-aisummary\" data-starttime=\"00:25:42\">00:25:42</a>. Audience members ask about the cost — around $2500 per NUC Pro system or $7000 for the full demo stack — and how workload scheduling occurs across distributed machines <a class=\"medius-aisummary\" data-starttime=\"00:26:11\">00:26:11</a>. Colin explains that models are sharded across three systems using Llama RPC servers, which handle distributed calls automatically <a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>. He notes that OpenVINO support for multi-device execution is under development but already excels on single systems <a class=\"medius-aisummary\" data-starttime=\"00:28:10\">00:28:10</a>. The session concludes with invitations to visit Intel’s booth for deeper hands-on experiences and thanks to the attendees <a class=\"medius-aisummary\" data-starttime=\"00:28:45\">00:28:45</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","speakerIds":["1768245294609build26-1778106036930001Ry3a","1768245294609build26-1778105890363001FusU","1768245294609build26-1777588199349001unfM","1768245294609build26-1779147084085001ynzm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.844Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Intel"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9ae79b2c-a2bb-44f2-80a7-dfd57d740b73","ogTitle":"Scale agentic AI from on-device to cloud orchestration","ogDescription":"Modern AI systems span client, edge, and cloud rather than a single model. Explore how agentic AI workloads operate across these environments through three demos: real‑time on‑device agents, distributed inference across edge systems, and enterprise‑scale multi‑agent orchestration on Azure Kubernetes Service with Intel Xeon. Leave with practical guidance on placing inference, reasoning, and orchestration where they make the most sense to balance responsiveness, scale, and efficiency.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777588199349001unfM","RegistrantKey":"1768245294609build26-1777588199349001unfM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778106036930001Ry3a","RegistrantKey":"1768245294609build26-1778106036930001Ry3a"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779147084085001ynzm","RegistrantKey":"1768245294609build26-1779147084085001ynzm"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778105890363001FusU","RegistrantKey":"1768245294609build26-1778105890363001FusU"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/BRKSP92_Intel_ScaleAI_ondevicetocloud_Build26_1780079348121001wkn5_1780512342157001b2EO.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"MS Teams","logicalValue":"MS Teams"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"AppTrust","logicalValue":"AppTrust"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Cross-device","logicalValue":"Cross-device"},{"displayValue":"App Integration","logicalValue":"App Integration"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"}]},{"@search.score":1.0,"documentId":"10489f4d-2740-40b0-ad56-5dc70683edd8","startDateTime":"2026-06-03T21:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mark Russinovich, Scott Hanselman, Darby Kosten","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701589001rRKd","langLocale":"en-US","sessionId":"1774018701589001rRKd","sessionInstanceId":"1774018701589001rRKd","sessionCode":"BRK247","sessionCodeNormalized":"EN-US-BRK247","title":"Scott and Mark learn...how agents reshape software engineering","sortTitle":"scott and mark learn...how agents reshape software engineering","description":"AI is changing how software is created—and what it means to be a software engineer. We’ll explore how AI agents are reshaping development: where they accelerate progress, where they fall short, and what’s changing for the profession. Along the way, we’ll share failure modes, lessons learned, and propose ways engineers and organizations can adapt. Real talk, no hype.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"AI agents are changing software engineering from prompt-by-prompt coding to handling larger workstreams and even producing full apps. Developers are seeing major productivity gains using AI-assisted coding, including faster delivery with small teams. AI works especially well for small personal tools, such as quick utilities built for one user. There is a difference between “vibe coding” and real software engineering meant for production, maintenance, and collaboration. AI can generate convincing but incorrect solutions, including bugs, poor fixes, and false confidence. High activity does not always mean real productivity—AI can create the illusion of progress while producing weak results. Experienced engineers benefit more from AI because they can recognize mistakes and guide the output effectively. Junior engineers may struggle with AI-generated code because they may not yet know how to judge its quality. Some technical problems remain genuinely hard, and AI alone is not enough without strong engineering skill. Companies still need to invest in early-career engineers or risk creating a future talent gap.","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:45 - 22:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:30:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/174a3901-070f-4a05-8a0d-4bf680418cc2","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/174a3901-070f-4a05-8a0d-4bf680418cc2","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/174a3901-070f-4a05-8a0d-4bf680418cc2","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/174a3901-070f-4a05-8a0d-4bf680418cc2","speakerIds":["1768245294609build26-1776631798007001nGQM","1768245294609build26-1769008130185001WAvh","1768245294609build26-1774976696107001NbL1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.456Z","heroSession":true,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/174a3901-070f-4a05-8a0d-4bf680418cc2","ogTitle":"Scott and Mark learn...how agents reshape software engineering","ogDescription":"AI is changing how software is created—and what it means to be a software engineer. We’ll explore how AI agents are reshaping development: where they accelerate progress, where they fall short, and what’s changing for the profession. Along the way, we’ll share failure modes, lessons learned, and propose ways engineers and organizations can adapt. Real talk, no hype.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769008130185001WAvh","RegistrantKey":"1768245294609build26-1769008130185001WAvh"},{"SpeakerType":"Session Support In-person, Session Owner","SpeakerId":"1768245294609build26-1774976696107001NbL1","RegistrantKey":"1768245294609build26-1774976696107001NbL1"},{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1776631798007001nGQM","RegistrantKey":"1768245294609build26-1776631798007001nGQM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK247","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"685e90db-9842-4d94-b260-1d7eadfa9293","startDateTime":"2026-06-03T20:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rob Ferguson, Daniel Han, Mark Saroufin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018713555001rcaV","langLocale":"en-US","sessionId":"1774018713555001rcaV","sessionInstanceId":"1774018713555001rcaV","sessionCode":"BRK234","sessionCodeNormalized":"EN-US-BRK234","title":"Shipping custom models at scale from fine-tuning to inference","sortTitle":"shipping custom models at scale from fine-tuning to inference","description":"Fine tuning, serving, and inference are core components of real world AI adoption. In this panel Rob Ferguson, Daniel Han (Unsloth), Mark Saroufim (Stealth Startup) and other practitioners discuss how teams are customizing models, taking them to production, and running them efficiently at scale. The conversation covers tradeoffs in fine tuning techniques, infrastructure choices for serving, optimizing inference cost and latency, and what’s working today for builders shipping AI powered products.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Speaker Backgrounds:</strong> The session opens with Rob Ferguson welcoming attendees at Microsoft Build <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. Rob introduces himself as Vice President of Technology at Fireworks AI, a managed service for AI inference and training on Microsoft Azure. He is joined by Daniel Han from Unsloth and Mark Saroufim from Core Automation, both experts in large-scale AI performance and optimization. Daniel discusses Unsloth’s growth with over 300 million downloads on Hugging Face and its focus on optimizing training speed and memory efficiency <a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>. Mark shares his experience as a PyTorch core maintainer and founder of GPU Mode, focusing on kernel programming and building automated labs for GPU systems optimization <a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>.</p>\n\n<p><strong>AI Optimization and Reinforcement Learning Infrastructure:</strong> Rob transitions to the topic of optimization in real AI systems <a class=\"medius-aisummary\" data-starttime=\"00:02:40\">00:02:40</a>. He highlights the importance of GPU efficiency and the collapsing gap between open models and closed APIs following DeepSeek R1’s RL post-training breakthroughs. The discussion clarifies the contrast between a “recipe” for RL and actual infrastructure needed to run it. Daniel and Mark explain how frameworks built for supervised fine-tuning struggle with RL loops that require continuous inference and gradient updates <a class=\"medius-aisummary\" data-starttime=\"00:05:26\">00:05:26</a>. Daniel introduces techniques like memory offloading, weight sharing with LoRA, and gradient checkpointing to reduce memory load and improve reproducibility <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>. Rob and Daniel emphasize deterministic reinforcement learning challenges where hardware inconsistencies can alter outcomes <a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>.</p>\n\n<p><strong>Reinforcement Learning Algorithms: PPO, GRPO, and RLVR:</strong> The speakers delve deeper into reinforcement learning history and algorithm design <a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>. Daniel traces the lineage from OpenAI’s PPO by John Shulman to DeepSeek’s GRPO, which simplified RL by removing separate reward and value models, instead normalizing generated data <a class=\"medius-aisummary\" data-starttime=\"00:11:47\">00:11:47</a>. He explains RLVR (Reinforcement Learning with Verifiable Rewards) and the power of scoring systems for training quality reasoning <a class=\"medius-aisummary\" data-starttime=\"00:12:31\">00:12:31</a>. Rob connects this to the broader idea of using scoring as a feedback signal for improvement in areas like video compression and modeling intelligence. Daniel notes that defining a correct reward function remains a major pain point for RL adoption <a class=\"medius-aisummary\" data-starttime=\"00:14:07\">00:14:07</a>, while Mark warns of automated systems leading to reward hacking and adversarial self-play behaviors <a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>. They reference AI “cheating” analogies, including the Volkswagen Dieselgate incident, to illustrate how models exploit weak verification setups <a class=\"medius-aisummary\" data-starttime=\"00:23:02\">00:23:02</a>.</p>\n\n<p><strong>LoRA Fine-tuning and Accessibility:</strong> The conversation shifts to LoRA (Low-Rank Adaptation) <a class=\"medius-aisummary\" data-starttime=\"00:24:01\">00:24:01</a>. Daniel explains LoRA as a technique for efficient model adaptation where only small low-rank matrices are trained instead of full model weights, drastically reducing compute cost and memory usage. He cites experiments showing that even rank “1” LoRA setups work remarkably well for certain domains like math and coding <a class=\"medius-aisummary\" data-starttime=\"00:25:30\">00:25:30</a>. Rob elaborates that such efficiency enables teams with modest hardware to perform post-training, democratizing AI experimentation <a class=\"medius-aisummary\" data-starttime=\"00:27:27\">00:27:27</a>. Daniel adds that with open-source tools from Unsloth and managed services from Fireworks, practitioners can make training feasible on local machines while combining optimizations from foundations like PyTorch <a class=\"medius-aisummary\" data-starttime=\"00:28:26\">00:28:26</a>.</p>\n\n<p><strong>Kernel-Level Optimization and Automation:</strong> Mark leads a deep technical walkthrough on GPU kernel optimizations <a class=\"medius-aisummary\" data-starttime=\"00:30:20\">00:30:20</a>. He highlights how fusion compilers such as TorchCompile and CUDA Graphs remove costly memory round trips and streamline operation launches. This, he says, constitutes 80% of modern AI performance improvements, while custom kernels add the last 20% “secret sauce” of fine-tuned speed and memory efficiency. He describes the rise of GPU Mode and kernel competitions educating engineers through self-written optimized code. Both Mark and Daniel touch on TorchCompile configurations, flags, and kernel fusion possibilities that remain hidden from users <a class=\"medius-aisummary\" data-starttime=\"00:35:35\">00:35:35</a>. Daniel predicts that AI may eventually auto-generate complex kernels such as Flash Attention 5 <a class=\"medius-aisummary\" data-starttime=\"00:37:20\">00:37:20</a>, though both emphasize human creativity remains essential for discovering novel solutions like gradient checkpointing <a class=\"medius-aisummary\" data-starttime=\"00:38:01\">00:38:01</a>.</p>\n\n<p><strong>Speculative Decoding, Agentic Systems, and Closing Insights:</strong> Toward the end <a class=\"medius-aisummary\" data-starttime=\"00:40:01\">00:40:01</a>, Rob and Mark evaluate speculative decoding as a transformative inference technique, converting slow sequential decoding to parallel “prefill”-like performance, though Daniel notes acceptance rates can drop when distributions shift <a class=\"medius-aisummary\" data-starttime=\"00:41:20\">00:41:20</a>. They then explore challenges of maintaining coherence in long agentic tasks and discuss softmax saturation issues in lengthy contexts <a class=\"medius-aisummary\" data-starttime=\"00:43:09\">00:43:09</a>. In conclusion, both share advice to AI builders: Mark suggests focusing on system bottlenecks beyond model inference, while Daniel encourages mastering algorithmic foundations such as flash attention and kernel math before engineering implementation <a class=\"medius-aisummary\" data-starttime=\"00:44:12\">00:44:12</a>. Rob closes by thanking attendees, urging them to explore Unsloth, Core Automation, and Fireworks AI <a class=\"medius-aisummary\" data-starttime=\"00:46:06\">00:46:06</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"20:30 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/990ffabd-eceb-4850-96a4-e40c4c2e7df8","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/990ffabd-eceb-4850-96a4-e40c4c2e7df8","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/990ffabd-eceb-4850-96a4-e40c4c2e7df8","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/990ffabd-eceb-4850-96a4-e40c4c2e7df8","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/990ffabd-eceb-4850-96a4-e40c4c2e7df8","speakerIds":["1768245294609build26-1774905889138001JtYZ","1768245294609build26-1775582336728001AON2","1768245294609build26-1775582375994001WvjJ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.854Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/990ffabd-eceb-4850-96a4-e40c4c2e7df8","ogTitle":"Shipping custom models at scale from fine-tuning to inference","ogDescription":"Fine tuning, serving, and inference are core components of real world AI adoption. In this panel Rob Ferguson, Daniel Han (Unsloth), Mark Saroufim (Stealth Startup) and other practitioners discuss how teams are customizing models, taking them to production, and running them efficiently at scale. The conversation covers tradeoffs in fine tuning techniques, infrastructure choices for serving, optimizing inference cost and latency, and what’s working today for builders shipping AI powered products.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774905889138001JtYZ","RegistrantKey":"1768245294609build26-1774905889138001JtYZ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775582336728001AON2","RegistrantKey":"1768245294609build26-1775582336728001AON2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775582375994001WvjJ","RegistrantKey":"1768245294609build26-1775582375994001WvjJ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"PyTorch","logicalValue":"PyTorch"},{"displayValue":"Personalization","logicalValue":"Personalization"}]},{"@search.score":1.0,"documentId":"0618e40a-ac9c-4d40-8cef-d6dbb5570095","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chip Huyen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018713380001rxTK","langLocale":"en-US","sessionId":"1774018713380001rxTK","sessionInstanceId":"1774018713380001rxTK","sessionCode":"BRK233","sessionCodeNormalized":"EN-US-BRK233","title":"Software Defensibility in the era of AI coding","sortTitle":"software defensibility in the era of ai coding","description":"If the cost of building software is approaching 0, is the value of building software is also approaching 0? If most software can be easily replicated by AI, how can we defend our products after we've built it? This talk discusses what have traditionally been considered moats for software products, why they aren't really moats, and why we should, as builders, continue building.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Background:</strong> At the beginning of the talk <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>, Chip Huyen greets the audience and introduces herself, sharing her professional history from NVIDIA to Snorkel AI and then founding and selling her own AI infrastructure startup. She explains that her current focus is robotics, before transitioning to the core topic—software defensibility in the age of AI. She reflects that both her strengths, writing and coding, are now easily automated, prompting her to ask what aspects of work can still be defended against automation <a class=\"medius-aisummary\" data-starttime=\"00:00:50\">00:00:50</a>. The opening sets the stage for a deep look at how AI is reshaping value creation and competition in software.</p>\n\n<p><strong>The Vanishing Cost of Software and Replication Challenges:</strong> Moving into the main discussion <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>, Chip explores the idea that the cost of software creation is approaching zero, thanks to rapid AI-assisted code generation. Companies that once tested 20 variations can now test thousands cheaply, raising the question of whether the value of software also approaches zero when building becomes frictionless. She shares an anecdote about launching a weekend project that quickly gained traction—only to be replicated by someone using Claude Code a day later <a class=\"medius-aisummary\" data-starttime=\"00:02:51\">00:02:51</a>. This experience highlights how easily AI can duplicate not only effort but also creative output, challenging traditional notions of defensibility and unique value in software products.</p>\n\n<p><strong>Identifying Moats in the AI Ecosystem:</strong> Chip then engages the audience in an interactive discussion about what constitutes a “moat” for software products in this new landscape <a class=\"medius-aisummary\" data-starttime=\"00:06:16\">00:06:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>. Attendees propose proprietary data, trust or branding, and customer service as potential defenses. She elaborates that proprietary data can easily be acquired by companies with funds—making money, rather than data, the true moat. Similarly, trust and branding can erode quickly when product improvements occur at exponential pace, as seen when ChatGPT faced rapid competition from Claude and DeepSeek <a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a>. Distribution advantages and expertise are considered next, with Chip noting that both can also be purchased or replicated. The only lasting advantage, she argues, may be momentum—moving faster than competitors, though that creates intense pressure for endless iteration.</p>\n\n<p><strong>Building Value Amid AI Expansion:</strong> In the next segment <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>, Chip contends that despite the bleak implications, there is still meaningful value in building software. She describes “career audits” and the idea of defensibility of oneself—how engineers can find skills AI cannot yet automate. She discusses the rise of long-tail problems AI frontier labs often overlook: niche challenges that affect a minority of users, such as underrepresented languages and latency issues in voice chatbots. Using an example from her Iranian colleague struggling with Farsi output from ChatGPT <a class=\"medius-aisummary\" data-starttime=\"00:16:30\">00:16:30</a>, Chip argues that unique cultural and contextual understanding can serve as a defensible strength. Builders should focus on problems “big enough to be profitable but not big enough for AI labs to chase,” such as nuanced preferences and regional variability.</p>\n\n<p><strong>Human–AI Interaction and Evolving Tooling:</strong> Transitioning to practical engineering implications <a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a>, she examines new workflows introduced by AI coding tools like Claude Code and Codex. She contrasts terminals’ deliberate complexity with IDE usability and imagines hybrid tools that blend interactive control and ease of use. Chip covers challenges of cloud-hosted agent persistence, code review dynamics on GitHub, and shifting artifacts—from human-written code to AI-generated instructions. She introduces the idea of adapting legacy workflows for human-AI collaboration, API-centric designs over GUI interfaces, and even re-architecting monolithic repositories into modular codebases to support AI-assisted development. She connects these software design trends with examples from robotics and urban infrastructure, like robots interacting through “streetlight APIs” to cross streets <a class=\"medius-aisummary\" data-starttime=\"00:30:51\">00:30:51</a>.</p>\n\n<p><strong>Robotics, Safety, and The Future of AI Agents:</strong> In her concluding section <a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:42:00\">00:42:00</a>, Chip extends the concept of digital agents to the physical world. She discusses reversibility failures—like Claude Code deleting her database—and the importance of safety in physical AI, especially robotics for elderly care. Using examples such as battery failures in Unitree G1 robots and one-wheel transport accidents, she highlights how physical AI agents introduce exposure to irreversible events and necessitate design for safety, self-charging, and reliability. She projects a future where AI agents expand from generating tokens to performing digital then physical actions, bridging coding, reasoning, and manipulation with environment understanding. The talk closes with excitement about upcoming robotics demonstrations and Chip’s vision of AI-enhanced physical systems still grounded by human purpose and defensible expertise <a class=\"medius-aisummary\" data-starttime=\"00:42:36\">00:42:36</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:00 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/6d282a65-465a-4585-a92b-ca528598b689","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/6d282a65-465a-4585-a92b-ca528598b689","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/6d282a65-465a-4585-a92b-ca528598b689","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/6d282a65-465a-4585-a92b-ca528598b689","speakerIds":["1768245294609build26-1775514714965001BQ8v"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.852Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/6d282a65-465a-4585-a92b-ca528598b689","ogTitle":"Software Defensibility in the era of AI coding","ogDescription":"If the cost of building software is approaching 0, is the value of building software is also approaching 0? If most software can be easily replicated by AI, how can we defend our products after we've built it? This talk discusses what have traditionally been considered moats for software products, why they aren't really moats, and why we should, as builders, continue building.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775514714965001BQ8v","RegistrantKey":"1768245294609build26-1775514714965001BQ8v"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Copilot","logicalValue":"Azure Copilot"},{"displayValue":"PowerShell","logicalValue":"PowerShell"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"80916ed1-6229-4d13-ac37-b7b5a56592f1","startDateTime":"2026-06-03T20:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alberto Martinez, Morris Novello","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774391405488001iTYM","langLocale":"en-US","sessionId":"1774391405488001iTYM","sessionInstanceId":"1774391405488001iTYM","sessionCode":"BRKSP90","sessionCodeNormalized":"EN-US-BRKSP90","title":"Stop routing docstrings to 70B models with on-device AI on Snapdragon","sortTitle":"stop routing docstrings to 70b models with on-device ai on snapdragon","description":"Your AI coding assistant calls a 70B+ cloud model just to add a docstring. Snapdragon X2 Elite’s 80 TOPS NPU changes that. In this session, build a three-tier inference routing architecture—on‑device (≤13B), on‑prem (14B–34B), and cloud (70B+)—cutting cloud tokens by 67%, latency by 70%, and keeping most code local. Includes routing logic, quantization trade‑offs, and a deployable classifier.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Opening and Introduction:</strong> The speaker, Alberto Martinez, begins the session with a light-hearted acknowledgment of the post-lunch lull among the audience (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:18\">00:00:18</a>). Introducing himself as the head of software strategy for Qualcomm’s compute business, he explains that his talk will be somewhat improvised, fitting for the creative “improv” venue (<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>). He emphasizes one of his core beliefs—predicting the future requires creating it—and frames the session as an exploration of the “mystery” behind agentic AI, orchestration, and token usage in large language models (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>).</p>\n\n<p><strong>AI Usage and Token Economics:</strong> Martinez transitions into the growing reliance on AI, asking participants to raise their hands if they use AI models for several hours daily (<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>). He humorously describes how both professionals and everyday users—like his own wife—are beginning to “run out of tokens,” underscoring how token consumption has become an everyday constraint (<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>). With increased agentic activity, token demand will multiply by 10 to 100 times according to Qualcomm’s leadership at Computex (<a class=\"medius-aisummary\" data-starttime=\"00:04:25\">00:04:25</a>). Martinez stresses that token economy is not just a backend cost problem—it affects consumers, power usage, latency, and sustainability (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>).</p>\n\n<p><strong>Problem Framing and Technical Context:</strong> He provocatively advises developers to “stop routing docstrings to seven-billion parameter models,” suggesting that routine code descriptions don’t require giant systems (<a class=\"medius-aisummary\" data-starttime=\"00:05:10\">00:05:10</a>). Martinez calls this wasteful both economically and energetically. He introduces an internal research analysis showing the math behind token expenditure and suggests a distributed three-tier model for improving efficiency—running some AI tasks locally, some on-prem, and only the most complex in the cloud (<a class=\"medius-aisummary\" data-starttime=\"00:10:50\">00:10:50</a>). The Snapdragon X Elite’s performance metrics are highlighted as enabling local computation with powerful efficiency and memory handling (<a class=\"medius-aisummary\" data-starttime=\"00:11:25\">00:11:25</a>).</p>\n\n<p><strong>Quantization and Efficiency Gains:</strong> Martinez explains the concept of quantization—converting from floating point to integer arithmetic—to drastically cut power and memory requirements while preserving accuracy (<a class=\"medius-aisummary\" data-starttime=\"00:12:10\">00:12:10</a>). He shows how integer formats can reduce model size and computational cost by 50%, fitting previously oversized models locally without quality loss. Research data demonstrate that 50–70% of computation and cost can be saved by token-efficient routing and quantized execution (<a class=\"medius-aisummary\" data-starttime=\"00:15:20\">00:15:20</a>). Using an example, he notes that $36,000 per developer per month in cloud spending can be cut by thousands when applying distributed task allocation and optimized model selection (<a class=\"medius-aisummary\" data-starttime=\"00:15:52\">00:15:52</a>).</p>\n\n<p><strong>Demonstration and Practical Example:</strong> To illustrate, he shows a recorded demo previously featured at Computex (<a class=\"medius-aisummary\" data-starttime=\"00:17:25\">00:17:25</a>). A hybrid system divides a web-design prompt into smaller sub-tasks—only the most complex (floating lantern animation) goes to a large cloud model, while others run locally. The generated results are identical, but cloud processing costs four times more tokens (<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>). Martinez points out this demonstrates that most AI-generated code—from text to multimedia—is essentially structured programming, and intelligently routing subtasks can yield 50–75% cost reduction for the same quality. This approach not only saves money but eases environmental load, avoiding unnecessary cloud computation (<a class=\"medius-aisummary\" data-starttime=\"00:23:19\">00:23:19</a>).</p>\n\n<p><strong>Architecture, Classifier Concept, and Closing Q&A:</strong> Expanding the system architecture, Martinez describes a three-tier compute structure where a “classifier” determines which tasks run locally and which escalate to the cloud based on complexity and entropy (<a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>). This classifier is envisioned as critical intellectual property that can evolve and learn to optimize task routing automatically. He calls on developers to create their own, noting that retroactive self-learning classifiers could transform orchestration efficiency. During Q&A (<a class=\"medius-aisummary\" data-starttime=\"00:34:55\">00:34:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:42:15\">00:42:15</a>), attendees ask about practical uses across organizations and AI-driven orchestration for video generation. Martinez explains fallback mechanisms—“abort and reclassify or cloud-submit”—paralleling CPU prefetch logic. Concluding, he encourages everyone to analyze their AI logs, track complexity by task, optimize classifiers weekly, and “save the planet along with some polar bears” through smarter local computation (<a class=\"medius-aisummary\" data-starttime=\"00:33:43\">00:33:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:42:15\">00:42:15</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"20:30 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/77723cdd-4ddd-4ede-8582-8ec8771be197","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/77723cdd-4ddd-4ede-8582-8ec8771be197","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/77723cdd-4ddd-4ede-8582-8ec8771be197","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/77723cdd-4ddd-4ede-8582-8ec8771be197","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/77723cdd-4ddd-4ede-8582-8ec8771be197","speakerIds":["1768245294609build26-1776712855440001Vttg","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.636Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/77723cdd-4ddd-4ede-8582-8ec8771be197","ogTitle":"Stop routing docstrings to 70B models with on-device AI on Snapdragon","ogDescription":"Your AI coding assistant calls a 70B+ cloud model just to add a docstring. Snapdragon X2 Elite’s 80 TOPS NPU changes that. In this session, build a three-tier inference routing architecture—on‑device (≤13B), on‑prem (14B–34B), and cloud (70B+)—cutting cloud tokens by 67%, latency by 70%, and keeping most code local. Includes routing logic, quantization trade‑offs, and a deployable classifier.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776712855440001Vttg","RegistrantKey":"1768245294609build26-1776712855440001Vttg"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Windows Development","logicalValue":"Windows Development"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"4bee6121-3f62-43de-97ed-26b433760de8","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Lachie Evenson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018704329001rnnz","langLocale":"en-US","sessionId":"1774018704329001rnnz","sessionInstanceId":"1774018704329001rnnz","sessionCode":"BRK222","sessionCodeNormalized":"EN-US-BRK222","title":"The honest practitioner's take on agentic AI on Kubernetes","sortTitle":"the honest practitioner's take on agentic ai on kubernetes","description":"Agentic AI workloads don't behave like normal services. They're stateful, bursty, multi-step, and often span more than a single cluster. Most teams figure this out the hard way. This session is for developers building agentic AI in production. We'll show you how to get the most out of Kubernetes for training, inference, and agent orchestration, covering purpose-built tooling, managed options, open-source inference at scale, and AI-assisted dev tools that support real-world AI operations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Opening and Session Overview:</strong> Lachlan Evenson greets the audience with high energy at <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a> and introduces the session titled \"The Honest Practitioner's Take on Agentic AI on Kubernetes.\" He shares his decade of experience with Kubernetes and sets expectations for a 45-minute journey covering three main parts: understanding why Kubernetes is a foundation for AI workloads, identifying what works and what’s challenging, and exploring future directions in agentic AI systems. He aims to provide tactical insights, demos, and templates attendees can take back to their teams. Before diving into technical content, Lachlan recounts the excitement he felt achieving scale with containers and compares that to his current enthusiasm for agentic AI systems running on Kubernetes—a new frontier of creative possibility <a class=\"medius-aisummary\" data-starttime=\"00:01:30\">00:01:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:20\">00:02:20</a>.</p>\n\n<p><strong>Why Kubernetes for AI and Understanding the Trade-offs:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:03:37\">00:03:37</a>, Lachlan explores why Kubernetes is often the right substrate for AI at scale. He contrasts fully managed AI platforms like Microsoft Foundry with self-managed Kubernetes setups on AKS, noting trade-offs between control and complexity. Managed platforms offer speed and simplicity, while Kubernetes provides openness, composability, and granular tuning—ideal for organizations requiring cost transparency, sovereignty, and multi-cloud flexibility. He compares AI workload needs to a \"project management iron triangle\" of quality, speed, and cost <a class=\"medius-aisummary\" data-starttime=\"00:06:01\">00:06:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:34\">00:07:34</a>, elaborating that Kubernetes allows fine-tuning of these priorities. However, with that flexibility comes operational overhead, which AI systems can potentially help manage. Finally, he defines different workload profiles—from teams requiring composability at scale to platform engineers prioritizing open governance—and details how Kubernetes primitives like namespaces, GPU scheduling, and declarative management empower AI development <a class=\"medius-aisummary\" data-starttime=\"00:08:00\">00:08:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>.</p>\n\n<p><strong>The AI Stack: From Inference to Training to Agentic Systems:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:10:49\">00:10:49</a> onward, Lachlan breaks down AI workloads into three categories—inference, training, and agentic systems—and how each fits onto Kubernetes. Inference workloads are short-lived and latency-sensitive, training jobs are long-running and throughput-bound, while agentic systems require stateful, multi-modal, long-context coordination. Kubernetes provides the base for all, but needs additional orchestration layers for scheduling, GPU utilization, scaling, and durable state management. At <a class=\"medius-aisummary\" data-starttime=\"00:14:24\">00:14:24</a>, he transitions into a layered model of the AI-on-Kubernetes stack, starting from the substrate and moving up to inference and serving with KAITO and AI Runway, then training and fine-tuning with Ray and Anyscale, and ending with agentic orchestration. The first product deep dive focuses on AI Runway and KAITO: AI Runway offers a Kubernetes-native interface for one-click model deployment; KAITO automates operational tuning and scaling. A live demo later (<a class=\"medius-aisummary\" data-starttime=\"00:19:35\">00:19:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:12\">00:23:12</a>) shows user Ralph deploying massive models like DeepSeek-R1, demonstrating automatic GPU fitting, cost estimation, and dynamic scaling—all accessible through an OpenAI-compatible API endpoint.</p>\n\n<p><strong>Training, Fine-tuning, and the Role of Ray and Anyscale:</strong> Starting at <a class=\"medius-aisummary\" data-starttime=\"00:23:32\">00:23:32</a>, the discussion shifts to training and fine-tuning. Lachlan introduces Ray, a distributed computation framework that complements AKS by providing distributed job scheduling, multi-node orchestration, and GPU partitioning. He announces the public preview of \"Anyscale on Azure\" (<a class=\"medius-aisummary\" data-starttime=\"00:25:15\">00:25:15</a>), a managed Ray service jointly developed with Anyscale, integrated with Azure’s identity and billing ecosystem. Demonstrating how users can create, authenticate, and manage clusters entirely within the Azure portal, he showcases a full notebook-to-production pipeline where a product catalog model is fine-tuned using Ray, metrics are visualized, and deployment endpoints are automatically generated. This workflow reduces months of setup to minutes, illustrating Kubernetes’ and Ray’s synergy for production-ready AI training pipelines. Observability is built-in, ensuring teams maintain full visibility into GPU utilization and job dependencies across nodes <a class=\"medius-aisummary\" data-starttime=\"00:34:20\">00:34:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:35:04\">00:35:04</a>.</p>\n\n<p><strong>Agentic Systems and the Kubernetes Agent Reference Stack:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:35:06\">00:35:06</a>, Lachlan dives into orchestrating AI agents—autonomous, multi-tool systems that extend automation beyond model serving. Using Skills, MCP, and OpenClaw as core examples, he explains that “skills” are reusable actions agents perform within a permissions framework, while MCP standardizes communication between agents and tools. Microsoft’s release of an AKS MCP server allows natural-language operations on Kubernetes, such as provisioning clusters directly via LLM interfaces. The newly released Kubernetes Agent Reference Stack (KARS) provides a sandboxed, secure environment for agent experimentation, leveraging Azure Linux, namespaces, and RBAC for isolation. A short demo (<a class=\"medius-aisummary\" data-starttime=\"00:38:50\">00:38:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:12\">00:41:12</a>) shows how users can launch OpenClaw—an open multi-agent orchestrator—on Kubernetes in minutes, establishing skills, security controls, and live conversational capabilities within the cluster.</p>\n\n<p><strong>Announcements, Real-World Cases, and Closing Takeaways:</strong> Near <a class=\"medius-aisummary\" data-starttime=\"00:42:12\">00:42:12</a>, Lachlan announces new AKS features: Automatic node pools for simplified scaling, Azure Container Linux for secure, minimal OS footprints, AKS on bare metal for virtualization-free performance, and Fleet for managing multi-cluster topologies. Real-world examples include the Royal Bank of Canada using KAITO in production to securely serve AI models with compliance boundaries, and Wayve using Anyscale on AKS to train autonomous driving systems that generalize globally. Internally, Microsoft also employs “AKS Claw,” an agentic SRE assistant, to assist engineers with operational troubleshooting. Lachlan concludes at <a class=\"medius-aisummary\" data-starttime=\"00:49:15\">00:49:15</a> with actionable guidance: start with Kubernetes as the base, add an agent layer via MCP and OpenClaw, mix managed and open-source tools like KAITO and Ray appropriately, and prepare for multi-cluster scaling using Fleet. He ends by thanking the audience, encouraging feedback, and expressing genuine excitement to see what practitioners build next in the evolving space of agentic AI atop Kubernetes.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"00:00 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/9466a393-4172-4f0a-9849-190dab1b2623","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9466a393-4172-4f0a-9849-190dab1b2623","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9466a393-4172-4f0a-9849-190dab1b2623","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9466a393-4172-4f0a-9849-190dab1b2623","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9466a393-4172-4f0a-9849-190dab1b2623","speakerIds":["1768245294609build26-1774284996815001pSYL"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9466a393-4172-4f0a-9849-190dab1b2623","ogTitle":"The honest practitioner's take on agentic AI on Kubernetes","ogDescription":"Agentic AI workloads don't behave like normal services. They're stateful, bursty, multi-step, and often span more than a single cluster. Most teams figure this out the hard way. This session is for developers building agentic AI in production. We'll show you how to get the most out of Kubernetes for training, inference, and agent orchestration, covering purpose-built tooling, managed options, open-source inference at scale, and AI-assisted dev tools that support real-world AI operations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996815001pSYL","RegistrantKey":"1768245294609build26-1774284996815001pSYL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK222","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"CP&D","logicalValue":"CP&D"}]},{"@search.score":1.0,"documentId":"7b63775a-b5b0-49e7-8936-30c902aa068a","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vivek Chauhan, Alicia Frame, Jetashree Ravi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774909601301001YwAb","langLocale":"en-US","sessionId":"1774909601301001YwAb","sessionInstanceId":"1774909601301001YwAb","sessionCode":"BRKSP91","sessionCodeNormalized":"EN-US-BRKSP91","title":"Turn foundation models into production AI on Microsoft Foundry","sortTitle":"turn foundation models into production ai on microsoft foundry","description":"Move beyond generic foundation models and learn how to build production‑ready, use case‑specific AI with optimized training and inference with Fireworks AI. See how developers integrate Fireworks with Microsoft Foundry to customize models, improve inference performance, and deploy at scale. Through a live demo and real‑world case study, you’ll learn practical patterns for reducing cost, latency, and time‑to‑production in enterprise AI systems.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction:</strong> The session opens with Vivek Chauhan greeting participants and checking audio setup (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>). He introduces his co-presenter Jed and a special guest Nico from Harvey AI, outlining that today’s discussion will focus on saving costs and building production-grade inference-optimized AI systems using open source models served via Fireworks AI and deployed on Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>). The introduction sets the tone by emphasizing practical enterprise strategies for leveraging open models, and Vivek then transitions into the main concepts that drive their AI perspective.</p>\n\n<p><strong>Core AI Principles and Fireworks Overview:</strong> Vivek highlights three core truths about the current AI landscape (<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>). First, he notes the narrowing performance gap between open models and closed frontier models, enabling enterprises to switch to open models without losing quality. Second, differentiation now relies on embedding proprietary data and fine-tuning models, not using generic solutions. Third, leading companies continuously optimize their AI stacks using Fireworks, finding efficiency through a “flywheel” of training, evaluation, and redeployment (<a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a>). Vivek describes Fireworks AI’s operations — processing over 30 trillion tokens daily and supporting more than 10,000 enterprises globally. Customers such as Cursor, Uber, and DoorDash leverage Fireworks for custom fine-tuning and inference performance gains (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>). He emphasizes the complexity of inference optimization, where Fireworks’ stack handles CUDA kernels, quantization, adaptive caching, and speculative decoding, freeing clients from heavy infrastructure management.</p>\n\n<p><strong>Platform Ecosystem and Training Capabilities:</strong> Vivek discusses how Fireworks empowers customers to “own their AI stack” (<a class=\"medius-aisummary\" data-starttime=\"00:04:12\">00:04:12</a>). The platform meets creators at various AI maturity stages, offering tools for non-technical product managers through a “training agent” that automates data preparation, evaluation, and fine-tuning, as well as managed training services for engineers (<a class=\"medius-aisummary\" data-starttime=\"00:05:08\">00:05:08</a>). Advanced users can use the training API for full customization with their own loss functions and reward settings. Vivek underscores the simplicity of one-click deployments, the ability to chain fine-tuned models, and continuous iteration through a training-inference loop — all supported with guaranteed numerical fidelity. This seamless workflow compresses development cycles and allows rapid iteration, which customers report improves productivity and innovation dramatically.</p>\n\n<p><strong>Demo and Foundry Integration Announcement:</strong> The next section presents practical demos of Fireworks tools and the major announcement of its availability on Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:14:42\">00:14:42</a>). Jed explains that Foundry users can now use Fireworks inference via a unified endpoint, contract, and platform journey. Four key customer benefits are cited: performance through GPU-level optimization, rapid model availability on release day, deep control and customizability, and major cost savings compared to closed models (<a class=\"medius-aisummary\" data-starttime=\"00:17:23\">00:17:23</a>). He details deployment modes — serverless “pay-as-you-go,” provisioned throughput for dedicated workloads, and bring-your-own-weights for proprietary models. Live demos illustrate browsing the Foundry catalog, selecting Fireworks models (identified by “FW” prefixes), configuring rate limits, and deploying models or custom weights easily within Azure. Users can integrate these models directly into applications or existing agents using APIs, making the system highly extensible and adaptable to real enterprise needs.</p>\n\n<p><strong>Customer Examples and Demonstrations:</strong> The presentation then showcases case studies from customers like UI Path, Bolt, and Motif (<a class=\"medius-aisummary\" data-starttime=\"00:21:07\">00:21:07</a>). Each sought better performance and economics through Fireworks integration — achieving higher quality and throughput than competing closed models. The demo walks through steps to upload custom models, set architectures, and deploy instances without restrictions in Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:24:00\">00:24:00</a>). Jed demonstrates how to connect Fireworks models directly into Azure agents, showing simple configuration screens where agents can be linked to open weight or custom models, along with external search tools or custom functions. This section provides visual proof of the integration’s ease-of-use, confirming that enterprises can deploy entire AI pipelines quickly while maintaining flexibility in scaling and model management.</p>\n\n<p><strong>Harvey AI Collaboration and Conclusion:</strong> The final portion features Nico from Harvey AI, who leads applied research and shares insights into their collaboration with Fireworks and Microsoft (<a class=\"medius-aisummary\" data-starttime=\"00:27:26\">00:27:26</a>). Harvey recently launched the “Legal Agent Benchmark” to evaluate AI agents’ performance on complex legal tasks. Nico describes their hybrid approach combining open models like GLM 5.1 with closed advisor models, creating significant cost reductions and improved accuracy — up to 2.4 times cheaper and 30% better in quality (<a class=\"medius-aisummary\" data-starttime=\"00:33:26\">00:33:26</a>). He underscores why open models matter: domain-specific fine-tuning improves legal precision, offers better latency control, and supports secure, fully auditable deployments. Harvey’s partnership with Fireworks and Foundry accelerates iteration cycles, enabling faster post-training experiments and deployment within Microsoft’s enterprise-grade environment. The session closes by encouraging every enterprise and startup using Azure to explore Fireworks models for building efficient, scalable, and secure AI applications (<a class=\"medius-aisummary\" data-starttime=\"00:40:11\">00:40:11</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"18:30 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/1ce61725-4e6d-40e0-902f-a17d1366824a","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/1ce61725-4e6d-40e0-902f-a17d1366824a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/1ce61725-4e6d-40e0-902f-a17d1366824a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/1ce61725-4e6d-40e0-902f-a17d1366824a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/1ce61725-4e6d-40e0-902f-a17d1366824a","speakerIds":["1768245294609build26-1776288992886001q5Ev","1768245294609build26-1772039473777001dTM9","1768245294609build26-1776288796753001kjcb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.843Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Fireworks AI"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/1ce61725-4e6d-40e0-902f-a17d1366824a","ogTitle":"Turn foundation models into production AI on Microsoft Foundry","ogDescription":"Move beyond generic foundation models and learn how to build production‑ready, use case‑specific AI with optimized training and inference with Fireworks AI. See how developers integrate Fireworks with Microsoft Foundry to customize models, improve inference performance, and deploy at scale. Through a live demo and real‑world case study, you’ll learn practical patterns for reducing cost, latency, and time‑to‑production in enterprise AI systems.\n\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776288992886001q5Ev","RegistrantKey":"1768245294609build26-1776288992886001q5Ev"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776288796753001kjcb","RegistrantKey":"1768245294609build26-1776288796753001kjcb"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/FInal%20External%20-%20Fireworks%20AI%20on%20Microsoft%20Foundry_%20Foundation%20Models%20to%20Production%20AI_1780435089594001oSaU%20%281%29_1780512199025001zSCg.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"d6c5647c-8487-4bbc-87be-c0691c1b0069","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ronak Chokshi, Joe Filcik, Maria Naggaga","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018700806001rmT7","langLocale":"en-US","sessionId":"1774018700806001rmT7","sessionInstanceId":"1774018700806001rmT7","sessionCode":"BRK242","sessionCodeNormalized":"EN-US-BRK242","title":"Turn your agents into action: Connect tools, APIs, and documents","sortTitle":"turn your agents into action: connect tools, apis, and documents","description":"Tired of searching for the right tool(s) for your agents? Seeing your context window constantly blowing out of proportion? We are introducing a new product category that allows developers to choose the right toolset depending upon the agents' context, the use-case and functional boundaries, enabling efficient use of tools, with built-in governance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction:</strong> The video opens with Maria Naggaga welcoming attendees to day two of Microsoft Build <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:21\">00:00:21</a>. She acknowledges the full house and a minor cold, then introduces herself as a Product Manager working on Microsoft Foundry. Joining her is Joe Filcik from content understanding. Maria outlines the session, titled “Turn Your Agents into Action with Tools, APIs, and Documents,” promising announcements, demos, and closing remarks <a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>. She begins by describing Microsoft’s approach to building an agent platform using GitHub and Foundry to host agent runtimes, optimize agents, and distribute tools securely across M365 <a class=\"medius-aisummary\" data-starttime=\"00:01:14\">00:01:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:30\">00:01:30</a>. Foundry Tools offer a hub for pre-built and third-party tools, focusing on scalability and centralized lifecycle governance <a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a>. </p>\n\n<p><strong>Challenges and Core Concepts:</strong> Maria transitions to the growing complexity of integrating tools in agentic systems <a class=\"medius-aisummary\" data-starttime=\"00:02:18\">00:02:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. She explains that while LLMs provide reasoning, tool calling enables agents to take action, driving business value <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:49\">00:03:49</a>. With a rapidly expanding ecosystem of skills, plugins, APIs, connectors, and new tool types, steering appropriate tool usage is critical <a class=\"medius-aisummary\" data-starttime=\"00:03:16\">00:03:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:23\">00:03:23</a>. Maria outlines three main challenges: tool sprawl, management, and security. Because these tools often touch production code and databases, robust governance is essential <a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:03\">00:04:03</a>. Foundry addresses these challenges through features such as the tool catalog for discovery and publishing, creation for curation, discovery for token efficiency, and governance for access control <a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>. </p>\n\n<p><strong>Tool Catalog and the Toolbox Concept:</strong> Maria highlights Foundry’s tool catalog as a centralized registry housing multiple tool types including OpenAPI, MCP, A2A, logic app connectors, skills, and plugins <a class=\"medius-aisummary\" data-starttime=\"00:05:41\">00:05:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:19\">00:06:19</a>. Developers can maintain private registries via Azure API Center for additional governance <a class=\"medius-aisummary\" data-starttime=\"00:06:46\">00:06:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>. She moves on to the problem of integrating multiple tools with varying identities and credential systems when designing complex agents, which can lead to hundreds of integrations and inefficient development <a class=\"medius-aisummary\" data-starttime=\"00:07:23\">00:07:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:11\">00:08:11</a>. Introducing “Toolbox,” Maria describes it as a reusable, managed, and consistent interface bundling multiple tools into one accessible unit <a class=\"medius-aisummary\" data-starttime=\"00:09:10\">00:09:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>. Toolboxes enable developers to build once, reuse across agents, and leverage a unified endpoint and authentication model. Four pillars—Build, Discover, Consume, and Govern—define this approach, offering lifecycle management and security monitoring <a class=\"medius-aisummary\" data-starttime=\"00:10:14\">00:10:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:23\">00:11:23</a>. </p>\n\n<p><strong>Live Demonstrations:</strong> Maria performs several live demos to showcase Toolbox functionality <a class=\"medius-aisummary\" data-starttime=\"00:12:32\">00:12:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>. She demonstrates the complexity of manual tool authentication and contrasts it with the simplicity of Toolbox configuration—reducing extensive auth code to a single URL. Additional features include Tool Search, which dynamically selects the correct tool from metadata instead of overloading the context window <a class=\"medius-aisummary\" data-starttime=\"00:13:02\">00:13:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:01\">00:15:01</a>. Maria then builds skills and integrates new browser automation using Playwright, followed by creating and managing toolboxes in Foundry <a class=\"medius-aisummary\" data-starttime=\"00:17:03\">00:17:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:04\">00:20:04</a>. Despite minor demo hiccups, the live testing validates endpoints and tool search functionality. She emphasizes that a unified endpoint now connects any hosted agent across Copilot, GitHub SDKs, and future Claude integrations <a class=\"medius-aisummary\" data-starttime=\"00:20:25\">00:20:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:12\">00:21:12</a>. </p>\n\n<p><strong>Content Understanding by Joe Filcik:</strong> Joe Filcik continues by introducing “Content Understanding,” addressing challenges of extracting information from documents, videos, and presentations <a class=\"medius-aisummary\" data-starttime=\"00:22:31\">00:22:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:00\">00:23:00</a>. He explains that the service cleans and structures multimodal input into JSON and markdown, already powering Foundry IQ and Microsoft 365 Copilot <a class=\"medius-aisummary\" data-starttime=\"00:23:30\">00:23:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:02\">00:25:02</a>. Real use cases include Wolters Kluwer and DataSnipper using the technology for tax, audit, and document automation. The pipeline comprises Parse, Classify, and Extract stages enabling high-fidelity content capture across formats <a class=\"medius-aisummary\" data-starttime=\"00:25:06\">00:25:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:26:09\">00:26:09</a>. Upcoming enhancements incorporate GPT-5-based model engines, wider file support, and agent frameworks like Langchain for automation <a class=\"medius-aisummary\" data-starttime=\"00:26:36\">00:26:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:01\">00:27:01</a>. Extraction improvements include grounded bounding box tracking, confidence scoring, and knowledge-based training scheduled for July <a class=\"medius-aisummary\" data-starttime=\"00:29:20\">00:29:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:30:31\">00:30:31</a>. </p>\n\n<p><strong>Demo of Agentic Extraction and Closing Remarks:</strong> Joe illustrates “agentic extraction” through recorded demos focusing on complex fiber optic failure analysis <a class=\"medius-aisummary\" data-starttime=\"00:32:01\">00:32:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:33\">00:33:33</a>. He shows how the new system reasons across documents and multimedia to identify causes, budgets, and structured context grounded in source evidence. Demonstrations compare standard parsing versus Content Understanding-enabled modes to highlight more accurate data retrieval in Foundry IQ ingestion and real-time agent workflows <a class=\"medius-aisummary\" data-starttime=\"00:35:01\">00:35:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:39:41\">00:39:41</a>. Enhanced classification analyzers allow businesses to encode custom extraction rules for structured business logic, yielding precise answers like technician identification and document completion statuses <a class=\"medius-aisummary\" data-starttime=\"00:39:43\">00:39:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:40:49\">00:40:49</a>. Maria returns to demonstrate agent integration with Foundry’s unified endpoint, showing how Tool Search optimizes runtime resource usage <a class=\"medius-aisummary\" data-starttime=\"00:41:12\">00:41:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:42:11\">00:42:11</a>. The session concludes with links to build demos on ai.azure.com and aka.ms/fibey and an invitation to join the Foundry Discord, followed by applause and closing acknowledgements <a class=\"medius-aisummary\" data-starttime=\"00:42:23\">00:42:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:17\">00:43:17</a>. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"23:00 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:45:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e70ca1c3-f52b-4bcb-bc80-6e6afc4f5537","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e70ca1c3-f52b-4bcb-bc80-6e6afc4f5537","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e70ca1c3-f52b-4bcb-bc80-6e6afc4f5537","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e70ca1c3-f52b-4bcb-bc80-6e6afc4f5537","speakerIds":["1768245294609build26-1751939918029001Ts21","1768245294609build26-1776459863938001ayf1","1768245294609build26-1775078342403001SpDz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.455Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e70ca1c3-f52b-4bcb-bc80-6e6afc4f5537","ogTitle":"Turn your agents into action: Connect tools, APIs, and documents","ogDescription":"Tired of searching for the right tool(s) for your agents? Seeing your context window constantly blowing out of proportion? We are introducing a new product category that allows developers to choose the right toolset depending upon the agents' context, the use-case and functional boundaries, enabling efficient use of tools, with built-in governance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 2","logicalValue":"Festival Pavilion, Breakout 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751939918029001Ts21","RegistrantKey":"1768245294609build26-1751939918029001Ts21"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775078342403001SpDz","RegistrantKey":"1768245294609build26-1775078342403001SpDz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776459863938001ayf1","RegistrantKey":"1768245294609build26-1776459863938001ayf1"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/BRK242","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"MCP","logicalValue":"MCP"}]},{"@search.score":1.0,"documentId":"394f14c2-79ab-4106-bdba-b4056e9a4bc5","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nish Anil, Hazem El-Hammamy, Jeff Fritz","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703960001rRT9","langLocale":"en-US","sessionId":"1774018703960001rRT9","sessionInstanceId":"1774018703960001rRT9","sessionCode":"BRK220","sessionCodeNormalized":"EN-US-BRK220","title":"Using AI tools to teach old apps new tricks","sortTitle":"using ai tools to teach old apps new tricks","description":"Modernizing apps isn’t just rewriting code—it’s untangling dependencies, tracing data flows, and making changes without breaking production.\u00A0In this session, we’ll show how you can use agentic AI to take on the hardest parts of modernization: analyzing large codebases, mapping dependencies, planning upgrades, refactoring safely, while doing it all at scale. With GitHub Copilot modernization capabilities, you can move from legacy complexity to modernized apps in days, not months.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"AI is transforming application modernization by reducing manual work in upgrades, migrations, refactoring, and security improvements. Modernization is a major priority for IT leaders, with the document citing that 94% of IT leaders view app modernization as a top investment area. Successful modernization requires three core pillars: scale, customization, and governance. Azure Copilot and GitHub Copilot together provide an end-to-end modernization solution, connecting IT planning with developer execution. Azure Copilot supports estate-wide modernization planning, including app discovery, dependency mapping, ROI analysis, and migration wave planning. GitHub Copilot Modernization helps developers with code assessments, framework upgrades, migration planning, refactoring, replatforming, and execution within tools like VS Code and CLI. The platform supports many legacy and modern technologies, including .NET, Java, Windows, Linux, VMware, Oracle, SQL Server, Sybase, and mainframe systems. Mainframe modernization can reverse-engineer COBOL/JCL applications into documentation, architecture, business logic, and data mappings, then transform them into modern Java applications. Custom skills, rule books, and command center features let organizations apply their own standards, governance policies, reusable patterns, and portfolio-level oversight. A major benefit is parallel AI agents working across multiple applications at once, leading to significant reductions in modernization time and effort.","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"16:00 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/800c83e5-8266-4657-914e-a8f682b4328e","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/800c83e5-8266-4657-914e-a8f682b4328e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/800c83e5-8266-4657-914e-a8f682b4328e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/800c83e5-8266-4657-914e-a8f682b4328e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/800c83e5-8266-4657-914e-a8f682b4328e","speakerIds":["1768245294609build26-1770924720596001wFnD","1768245294609build26-1777576114739001uRBc","1768245294609build26-1768325213362001z95b"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.232Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/800c83e5-8266-4657-914e-a8f682b4328e","ogTitle":"Using AI tools to teach old apps new tricks","ogDescription":"Modernizing apps isn’t just rewriting code—it’s untangling dependencies, tracing data flows, and making changes without breaking production.\u00A0In this session, we’ll show how you can use agentic AI to take on the hardest parts of modernization: analyzing large codebases, mapping dependencies, planning upgrades, refactoring safely, while doing it all at scale. With GitHub Copilot modernization capabilities, you can move from legacy complexity to modernized apps in days, not months.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770924720596001wFnD","RegistrantKey":"1768245294609build26-1770924720596001wFnD"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777576114739001uRBc","RegistrantKey":"1768245294609build26-1777576114739001uRBc"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768325213362001z95b","RegistrantKey":"1768245294609build26-1768325213362001z95b"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"App Mod","logicalValue":"App Mod"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Reserve","logicalValue":"Reserve"}]},{"@search.score":1.0,"documentId":"86d8dfa2-2838-4caa-a56b-fc51b2781980","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pierce Boggan, Josh Spicer, Anna Soracco","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708919001rKEA","langLocale":"en-US","sessionId":"1774018708919001rKEA","sessionInstanceId":"1774018708919001rKEA","sessionCode":"BRK204","sessionCodeNormalized":"EN-US-BRK204","title":"What we learned shipping VS Code weekly (without breaking everything)","sortTitle":"what we learned shipping vs code weekly (without breaking everything)","description":"Shipping faster sounds great until test gaps, review bottlenecks, and triage backlog scale with it. The VS Code team hit all of that going from monthly to weekly releases, and agents made it work. This session breaks down the real patterns: agent sessions before meetings, conversations that become PRs instead of specs, automated triage across one of GitHub's largest repos, and harnesses that keep quality high when 100+ commits land daily. Concrete workflows you can take back to your team.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The video opens with Pierce Boggan from the VS Code team welcoming viewers and explaining the focus of the session — what the team learned when transitioning from monthly to weekly releases <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>. Instead of unveiling new AI products as other Build sessions did, this talk concentrates on the VS Code team’s internal evolution while adopting AI technologies over the past year and a half. Pierce and his colleague Josh Spicer introduce themselves and emphasize how their work now integrates AI deeply into engineering, planning processes, and product delivery <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>. They set the stage by describing how VS Code progressed from a simple editor into an AI-driven platform supported by agents like Copilot, making development more rapid but also complex <a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a>. The introduction closes by showing metrics such as the \"code survival rate\" for AI-generated commits and how agent-based coding increased productivity and issues simultaneously, prompting the move to weekly releases in early 2024 <a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:40\">00:04:40</a>.</p>\n\n<p><strong>Adopting Weekly Release Cycles and Implications of AI Adoption:</strong> Pierce narrates the team’s growing pains and opportunities after successfully integrating AI into internal workflows <a class=\"medius-aisummary\" data-starttime=\"00:04:40\">00:04:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>. He explains that the pace of AI innovation and competitive pressure required moving from monthly to weekly releases, starting in February 2024. This shift was not only about releasing faster features but about handling increased issues and pull requests generated by both the team and the open-source community. Weekly cadence helped limit risks associated with huge monthly code drops, prevented accumulation of deployment freezes during holidays, and ensured smoother incremental validation <a class=\"medius-aisummary\" data-starttime=\"00:05:15\">00:05:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:54\">00:05:54</a>. Pierce emphasizes that success with AI means rethinking engineering systems, planning, and how teams collaborate efficiently under compressed innovation timelines <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>.</p>\n\n<p><strong>Developer Inner Loop and AI-Enhanced Tools:</strong> Josh Spicer demonstrates the AI-centric tools that transformed their development workflow <a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:45\">00:14:45</a>. He illustrates a development scenario using the \"component browser\" tool — vital for visually inspecting and testing UI elements without compiling full builds. This tool allows agents to inject and validate code changes, automate screenshots of before-and-after states, and catch regressions inside GitHub pull requests <a class=\"medius-aisummary\" data-starttime=\"00:07:27\">00:07:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>. Josh demonstrates collaborative prototyping workflows powered by Copilot agents, where PMs and engineers ideate interactively within VS Code. AI helps synchronize development between contributors without relying solely on text-based specs, making real-time experimentation possible <a class=\"medius-aisummary\" data-starttime=\"00:12:30\">00:12:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:45\">00:13:45</a>. He also shares about new agentic skills such as \"chat performance skill,\" which automatically benchmarks and validates performance after UI or code changes through built-in telemetry. These advancements streamline testing, reduce regressions, and make performance intelligence accessible across the team <a class=\"medius-aisummary\" data-starttime=\"00:15:11\">00:15:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:25\">00:18:25</a>.</p>\n\n<p><strong>Model Integration and Evaluation Systems:</strong> Pierce returns to explain the complex behind-the-scenes process of integrating new AI models into VS Code <a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:00\">00:24:00</a>. New releases from model providers, such as OpenAI or Anthropic, require weeks of prompt optimization and offline testing before public rollout. He describes the \"Copilot API\" gateway and the internal \"VSC-Bench\" evaluation stack that runs thousands of tests daily to optimize reasoning effort, token usage, and resolution rates. Offline results are complemented by live A/B experiments with telemetry-enabled VS Code users for real-world validation. The team carefully balances quality and cost trade-offs, adjusting prompt structure dynamically for maximum efficiency <a class=\"medius-aisummary\" data-starttime=\"00:21:19\">00:21:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:40\">00:23:40</a>. These systems ensure AI-infused features are not only powerful but sustainable and measurable in production environments.</p>\n\n<p><strong>AI-Native Engineering System and Automated Issue Management:</strong> Josh presents the engineering system upgrades that allow VS Code to handle an unprecedented volume of issues and contributions <a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:37:00\">00:37:00</a>. Their bespoke automation architecture connects GitHub events, telemetry data, and scheduled agents through a resilient pipeline. He demonstrates how auto-triaging identifies issue owners via data from commits and ownership mappings, and how AI deduplicates related issues using semantic understanding rather than simple keyword matching <a class=\"medius-aisummary\" data-starttime=\"00:29:56\">00:29:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:32:11\">00:32:11</a>. Josh also walks through an example where the team’s error agent autonomously detects unhandled exceptions, generates GitHub issues, and proposes fixes — showing an actual regression resolved automatically within days <a class=\"medius-aisummary\" data-starttime=\"00:35:00\">00:35:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:43\">00:36:43</a>. Together with incremental and staged release rollouts <a class=\"medius-aisummary\" data-starttime=\"00:38:02\">00:38:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:39:05\">00:39:05</a>, these automations safeguard the product as it adapts weekly, ensuring stability while maintaining developer velocity.</p>\n\n<p><strong>Collaboration Model Evolution and Conclusion:</strong> In closing, Pierce details how VS Code’s planning and collaboration model evolved alongside its AI adoption <a class=\"medius-aisummary\" data-starttime=\"00:40:37\">00:40:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:45:06\">00:45:06</a>. Traditional quarterly or even monthly planning cycles proved obsolete given rapid AI-driven delivery. The team now operates through dynamic, temporary \"workstreams\" with designated DRIs (Directly Responsible Individuals) focused on high-priority goals for short spans. These small agile groups plan, execute, and disband as needs shift, allowing adaptation without losing alignment across the 40-person team <a class=\"medius-aisummary\" data-starttime=\"00:42:33\">00:42:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:44\">00:43:44</a>. Pierce emphasizes that successful AI integration forces organizations to reexamine everything — process, planning, engineering stability, and delivery cadence. The presenters encourage listeners to rethink their workflows similarly and invite conversations after the session, closing on a note of collaboration and shared learning as applause fills the room <a class=\"medius-aisummary\" data-starttime=\"00:45:02\">00:45:02</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/8e6c2160-2e93-40ac-8b17-a731909983b0","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8e6c2160-2e93-40ac-8b17-a731909983b0","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8e6c2160-2e93-40ac-8b17-a731909983b0","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8e6c2160-2e93-40ac-8b17-a731909983b0","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8e6c2160-2e93-40ac-8b17-a731909983b0","speakerIds":["1768245294609build26-1775704791363001hLZq","1768245294609build26-1774285000548001pBWh","1768245294609build26-1774284986630001pTM8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.466Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8e6c2160-2e93-40ac-8b17-a731909983b0","ogTitle":"What we learned shipping VS Code weekly (without breaking everything)","ogDescription":"Shipping faster sounds great until test gaps, review bottlenecks, and triage backlog scale with it. The VS Code team hit all of that going from monthly to weekly releases, and agents made it work. This session breaks down the real patterns: agent sessions before meetings, conversations that become PRs instead of specs, automated triage across one of GitHub's largest repos, and harnesses that keep quality high when 100+ commits land daily. Concrete workflows you can take back to your team.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Session Support In-person","SpeakerId":"1768245294609build26-1774284986630001pTM8","RegistrantKey":"1768245294609build26-1774284986630001pTM8"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775704791363001hLZq","RegistrantKey":"1768245294609build26-1775704791363001hLZq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000548001pBWh","RegistrantKey":"1768245294609build26-1774285000548001pBWh"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK204","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"145b748f-457c-4e7c-88d8-61377e004fae","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mario Rodriguez, Evan Boyle","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708247001rStk","langLocale":"en-US","sessionId":"1774018708247001rStk","sessionInstanceId":"1774018708247001rStk","sessionCode":"BRK200","sessionCodeNormalized":"EN-US-BRK200","title":"Why your AI code doesn’t ship: Closing the gap to production","sortTitle":"why your ai code doesn’t ship: closing the gap to production","description":"AI isn't just autocompleting your code anymore — it's writing plans, shipping PRs, fixing pipelines, and patching prod. In this demo-heavy session, we'll show AI agents working across the entire dev lifecycle: planning, coding, CI/CD, and live operations. You'll see how to move faster, keep agents on a leash, and build systems that fix themselves.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Opening and Vision for Agents:</strong> Mario Rodriguez begins by welcoming the audience to Build 2026 and introducing himself as the leader of product at GitHub (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>). He sets a tone of enthusiasm for developer creativity and declares that the session will be about \"demos, not memos.\" He outlines a key shift occurring in the software world: the entry of AI agents into the workforce (<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>). Unlike older automation systems requiring constant supervision, modern agents allow for \"macro-delegation and micro-steering,\" where developers set goals instead of micromanaging steps. This change redefines workflows, merges the traditional software development lifecycle with product development, and emphasizes collaboration between humans and agents. Mario asserts that velocity and creativity will drastically increase as developers focus on judgment and imagination with agents supporting execution.</p>\n\n<p><strong>Defining the Agent-Native System:</strong> Mario transitions into GitHub’s evolution toward becoming an “agent-native engineering system” (<a class=\"medius-aisummary\" data-starttime=\"00:05:51\">00:05:51</a>). He lays out the platform’s foundational pillars: multiple work surfaces (IDEs, command line, GitHub Copilot app), secure runtimes both local and cloud-based, automation for predictable scale, quality verification systems, continuous learning memory, and enterprise-grade trust. By uniting these, GitHub is building an API- and UX-first platform built for human-agent collaboration. The goal is to help developers delegate confidently while maintaining high code quality and safety. Before concluding his segment, Mario invites Evan Boyle to the stage (<a class=\"medius-aisummary\" data-starttime=\"00:07:17\">00:07:17</a>)—the co-creator of the Copilot CLI, SDK, and new Copilot app—to demonstrate these concepts in practice.</p>\n\n<p><strong>Demonstrating the GitHub Copilot App and Canvases:</strong> Evan introduces the GitHub Copilot app not as another task manager but as a focus tool for finishing work rather than multiplying it (<a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>). Through live demos, he shows how engineers can batch issues, launch agentic sessions, and delegate debugging or feature work directly from the interface. Evan demonstrates “canvases”—interactive, customizable panels the agent can manipulate and even create dynamically (<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>). He builds and launches a “whiteboard canvas” that allows real-time drawing and publishing to a GitHub gist within minutes. The feature enables deep team-customized workflows, with canvases also integrated from partners like Miro. It showcases a shift from static tools to living, extensible surfaces driven bi-directionally by agents and users.</p>\n\n<p><strong>Expanding into Cloud, Chronicle, and Agent Merge:</strong> Evan moves on to illustrate new layers built into the Copilot ecosystem. “Copilot Cloud,” powered by GitHub Sandboxes, allows developers to offload live sessions to secure micro-VMs (<a class=\"medius-aisummary\" data-starttime=\"00:21:12\">00:21:12</a>)—letting agents continue tasks while developers disconnect. “Chronicle” acts as a unified, private memory system tracking every coding session, bug fix, and prompt across devices. It can summarize work for daily standups using simple commands like “/chronicle standup” to generate structured updates automatically (<a class=\"medius-aisummary\" data-starttime=\"00:24:44\">00:24:44</a>). Finally, Evan unveils “Agent Merge,” which autonomously manages pull requests, reacts to feedback, resolves conflicts, and merges code once tests pass (<a class=\"medius-aisummary\" data-starttime=\"00:26:07\">00:26:07</a>). This automation closes the loop between writing, reviewing, and delivering production-ready results, shifting the focus back to valuable outcomes rather than activity noise.</p>\n\n<p><strong>Automation, Quality, and Security Reviews:</strong> Returning to the stage together, Mario and Evan emphasize automation as essential for both speed and trustworthiness. Evan presents “Morning Briefs”—personal automations summarizing the day’s critical issues or in-flight PRs each morning (<a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a>). He follows with a “cloud automation” that attempts to reproduce new bugs automatically and comment back with reproducible test cases to accelerate fixes (<a class=\"medius-aisummary\" data-starttime=\"00:35:01\">00:35:01</a>). They then demonstrate quality control through multi-agent and multi-model code and security reviews that use several large models such as Gemini, GPT, and Claude working in parallel (<a class=\"medius-aisummary\" data-starttime=\"00:41:01\">00:41:01</a>). Integrated “Rubber Duck” critique sessions ensure output validation before human attention is required, producing high-confidence results with minimal redundancy or error risk.</p>\n\n<p><strong>Conclusion and Vision Forward:</strong> In closing, Mario reiterates that GitHub’s next phase is about moving from AI assistance to true human-agent collaboration (<a class=\"medius-aisummary\" data-starttime=\"00:43:38\">00:43:38</a>). The platform’s new “agent-native engineering system” spans code generation, automation, runtime safety, memory continuity, and enterprise-grade orchestration. By partnering with technologies like Sonar, Miro, Endor, and LaunchDarkly, GitHub ensures security and scalability across every development stage. The presenters end with the whiteboard canvas functioning perfectly and celebrate it live (<a class=\"medius-aisummary\" data-starttime=\"00:30:00\">00:30:00</a>), illustrating how imagination can become real through agents. The session closes with an invitation to developers everywhere to embrace this vision—building software alongside intelligent partners that extend human focus, creativity, and impact.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"19:30 - 20:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/a6e942b5-ceab-40f1-9a69-03390a58d306","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a6e942b5-ceab-40f1-9a69-03390a58d306","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a6e942b5-ceab-40f1-9a69-03390a58d306","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a6e942b5-ceab-40f1-9a69-03390a58d306","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a6e942b5-ceab-40f1-9a69-03390a58d306","speakerIds":["1768245294609build26-1775133280383001hbCQ","1768245294609build26-1776135903186001Orl7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a6e942b5-ceab-40f1-9a69-03390a58d306","ogTitle":"Why your AI code doesn’t ship: Closing the gap to production","ogDescription":"AI isn't just autocompleting your code anymore — it's writing plans, shipping PRs, fixing pipelines, and patching prod. In this demo-heavy session, we'll show AI agents working across the entire dev lifecycle: planning, coding, CI/CD, and live operations. You'll see how to move faster, keep agents on a leash, and build systems that fix themselves.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Cowell Theater","logicalValue":"Gateway Pavilion, Level 1, Cowell Theater"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775133280383001hbCQ","RegistrantKey":"1768245294609build26-1775133280383001hbCQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776135903186001Orl7","RegistrantKey":"1768245294609build26-1776135903186001Orl7"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/VSCode/Learn?utm_source=build-brk201-related-vscode-learn-cta&utm_medium=event&utm_campaign=msbuild-2026","backpackLink":null,"label":"Learn agent-first development in VS Code","category":"Web link"},{"link":"https://aka.ms/VSCode/Live?utm_source=build-brk201-related-vscode-live-cta&utm_medium=event&utm_campaign=msbuild-2026","backpackLink":null,"label":"Sign up for our next Release livestream (ask follow up Qs)","category":"Web link"},{"link":"https://aka.ms/AwesomeCopilot?utm_source=build-brk201-related-awesome-copilot-cta&utm_medium=event&utm_campaign=msbuild-2026","backpackLink":null,"label":"Community-contributed agents, instructions, and skills","category":"Web link"}],"nextStep":{"link":"https://aka.ms/VSCode/Learn?utm_source=build-brk201-next-vscode-learn-cta&utm_medium=event&utm_campaign=msbuild-2026","backpackLink":null,"label":"Learn agent-first development in VS Code","category":"Web link"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"1d2feb7f-57ed-4d3d-b1b4-e9bd5a8d8019","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mario Rodriguez, Evan Boyle","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778109100437001YTWq","langLocale":"en-US","sessionId":"1778109100437001YTWq","sessionInstanceId":"1778109100437001YTWq","sessionCode":"BRK200-R1","sessionCodeNormalized":"EN-US-BRK200-R1","title":"Why your AI code doesn’t ship: Closing the gap to production","sortTitle":"why your ai code doesn’t ship: closing the gap to production","description":"AI isn't just autocompleting your code anymore — it's writing plans, shipping PRs, fixing pipelines, and patching prod. In this demo-heavy session, we'll show AI agents working across the entire dev lifecycle: planning, coding, CI/CD, and live operations. You'll see how to move faster, keep agents on a leash, and build systems that fix themselves.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Opening Remarks and Setting the Stage:</strong> The session begins with Mario Rodriguez greeting the audience and setting an informal tone for this end-of-conference talk (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>). He reflects on presenting the same content earlier after the keynote and now using it to close out the Build conference. Mario revisits a key development from December, where AI models in coding significantly improved, enabling a shift from constant correction to true “macro-delegation.” He compares early AI model management to supervising a toddler—requiring constant oversight—but as models matured, developers began “micro-steering” instead of micromanaging agents. By January, GitHub saw a surge in agent activity that even maxed out data center capacity, leading to offloading into Azure (<a class=\"medius-aisummary\" data-starttime=\"00:02:04\">00:02:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>). Mario emphasizes this moment as the entry of “agents into the workforce,” envisioning GitHub’s evolution from human-only collaboration to a hybrid human-agent environment. He predicts that the familiar software development lifecycle (SDLC) will become unrecognizable within five years, introducing new workflow bottlenecks in areas like pull request validation.</p>\n\n<p><strong>Foundations of an Agent-Native Engineering System:</strong> Mario details the architectural pillars for GitHub’s agent-native future (<a class=\"medius-aisummary\" data-starttime=\"00:03:46\">00:03:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:38\">00:08:38</a>). First, he discusses “new surfaces” beyond traditional IDEs—ushering in Agentic Development Environments (ADEs) that move past terminal interfaces and integrate directly with agent workflows. Second, he highlights the emergence of new run-time primitives such as secure sandboxes with built-in tools optimized for agentic automation. Third, he likens AI automation to 3D printing, arguing that developers should focus creative energy on design (“the CAD”) while automation reliably executes standard tasks. To ensure sustainable quality, GitHub is focusing on “AI quality control,” addressing issues like security, maintainability, and architectural soundness—especially since current models struggle with architecture-level reasoning. Finally, he outlines the need for enterprise-grade trust mechanisms—observability, cost tracking (FinOps), and memory systems that let agents, like humans, “learn” iteratively to improve productivity and reliability across the SDLC.</p>\n\n<p><strong>Lowering Barriers and Raising Ceilings in Software Creation:</strong> Transitioning toward practical impact, Mario introduces the freshly announced GitHub Copilot App (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:01\">00:11:01</a>). In collaboration with Head of Design John Maeda, he explains a philosophy of “lowering the floor and raising the ceiling”—expanding access to coding while empowering experts to achieve more ambitious results. Lowering the floor makes development accessible to new creators, potentially unlocking “a billion developers,” while raising the ceiling equips professionals with tools to push software capabilities beyond current limits. He compares this democratization to making musical instruments accessible to more potential “Mozarts.” At this point, Mario invites engineering manager Evan Boyle to demo the Copilot App—emphasizing GitHub’s “demos, not memos” culture and encouraging live, unscripted experimentation.</p>\n\n<p><strong>Copilot App Demos and Personalized Agent Workflows:</strong> Evan begins his demo by presenting the new GitHub Copilot App as a workflow-finishing tool rather than just a multitask environment (<a class=\"medius-aisummary\" data-starttime=\"00:12:04\">00:12:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>). He shows how developers can manage issues, start sessions directly from issue contexts, and automate tasks through voice commands and canvases—custom, agent-driven UIs embedded in the app. Through interactive “triage board” and “Kanban board” examples, he demonstrates bi-directional communication between agents and UI elements, real-time analysis of GitHub issues, and dynamic canvas creation using the “Create Canvas” skill. Developers can share these canvases via GitHub gists or repositories, creating reusable interactive tools. Evan showcases collaborative extensions, such as Miro integration, and highlights the marketplace for community and partner-built canvases. These capabilities allow individuals and teams to operate with personalized, agent-assisted workflows, shifting the developer’s focus from repetitive execution to higher-level problem solving.</p>\n\n<p><strong>Continuity, Chronicle, and Automation at Scale:</strong> Building on those demos, Evan explores “continuity” features that enable sustained development across devices and sessions (<a class=\"medius-aisummary\" data-starttime=\"00:26:00\">00:26:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:32:00\">00:32:00</a>). He introduces GitHub Sandboxes—micro-VMs that retain context between sessions and suspend during inactivity—allowing tasks to run while developers step away. With the new cloud-based CLI mode, engineers can delegate long builds or tests overnight and resume seamlessly later. Next, he highlights “Chronicle,” a private memory system indexing data from all Copilot sessions and devices, enabling retrospective queries like “find that tricky Rust bug from last week.” Chronicle can automatically generate team standups, extracting updates from recent commits or sessions to summarize daily progress. This automation embodies GitHub’s vision of using AI to reclaim developer attention, focusing effort on meaningful outcomes rather than status management.</p>\n\n<p><strong>Security, Quality Reviews, and Closing Vision:</strong> Returning to advanced collaboration, Evan demonstrates the Copilot App’s PR and security review features (<a class=\"medius-aisummary\" data-starttime=\"00:40:00\">00:40:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:47:30\">00:47:30</a>). Using nested agent sessions, he runs parallel code and security reviews via multiple AI models—Sonnet, GPT, Gemini—then reconciles them through “Rubber Duck,” an internal critiquing agent that merges analysis for precision. This multilayered process heightens code quality while protecting human focus by automating repetitive checks. Evan also showcases daily “morning briefs” and automated bug reproducers that independently verify issue reports and generate failing test cases. As the talk concludes (<a class=\"medius-aisummary\" data-starttime=\"00:48:01\">00:48:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:48:44\">00:48:44</a>), Mario summarizes the concept of an “agent-native engineering system”—spanning surfaces, runtime, automation, quality, enterprise trust, and memory—and thanks the audience for embracing this next stage of human-agent software creation. Evan closes by inviting everyone to download the GitHub App from GitHub.com/app and experiment with building their own agentic workflows.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"23:00 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:45:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/7d0f35c8-cd91-4e97-ad44-9161a7ad7cf8","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/7d0f35c8-cd91-4e97-ad44-9161a7ad7cf8","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/7d0f35c8-cd91-4e97-ad44-9161a7ad7cf8","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/7d0f35c8-cd91-4e97-ad44-9161a7ad7cf8","speakerIds":["1768245294609build26-1775133280383001hbCQ","1768245294609build26-1776135903186001Orl7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.769Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/7d0f35c8-cd91-4e97-ad44-9161a7ad7cf8","ogTitle":"Why your AI code doesn’t ship: Closing the gap to production","ogDescription":"AI isn't just autocompleting your code anymore — it's writing plans, shipping PRs, fixing pipelines, and patching prod. In this demo-heavy session, we'll show AI agents working across the entire dev lifecycle: planning, coding, CI/CD, and live operations. You'll see how to move faster, keep agents on a leash, and build systems that fix themselves.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building B, Level 3, BATS Improv","logicalValue":"Building B, Level 3, BATS Improv"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1775133280383001hbCQ","RegistrantKey":"1768245294609build26-1775133280383001hbCQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776135903186001Orl7","RegistrantKey":"1768245294609build26-1776135903186001Orl7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"c46807db-d391-4d21-8358-03adff2fb100","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Patrick Nikoletich, Steve Sanderson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709261001rXXz","langLocale":"en-US","sessionId":"1774018709261001rXXz","sessionInstanceId":"1774018709261001rXXz","sessionCode":"BRK206","sessionCodeNormalized":"EN-US-BRK206","title":"Your agent, anywhere: MultiClient, MultiDevice with GitHub Copilot SDK","sortTitle":"your agent, anywhere: multiclient, multidevice with github copilot sdk","description":"Agents are powerful on your machine, but what happens when you need them everywhere else? In this session, we'll show how GitHub Copilot SDK lets you build an agent, embed it in an app, and take it with you across devices and into the cloud. You'll see how to go from a local agent to one you can access on your phone, move between machines, and run across multiple clients. If you've been working with agents locally and wondering what the next step looks like, this is it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"21:30 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:15:00Z","durationInMinutes":45,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/a890cd3f-7018-44c2-8be3-de0469d36b1e","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/77be1034-59af-4388-bdb6-37a99f2acae7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/77be1034-59af-4388-bdb6-37a99f2acae7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/77be1034-59af-4388-bdb6-37a99f2acae7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/77be1034-59af-4388-bdb6-37a99f2acae7","speakerIds":["1768245294609build26-1777003136185001uvnL","1768245294609build26-1776185877803001f40v"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.787Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018709261001rXXz_20260512022044230.avif","ogTitle":"Your agent, anywhere: MultiClient, MultiDevice with GitHub Copilot SDK","ogDescription":"Agents are powerful on your machine, but what happens when you need them everywhere else? In this session, we'll show how GitHub Copilot SDK lets you build an agent, embed it in an app, and take it with you across devices and into the cloud. You'll see how to go from a local agent to one you can access on your phone, move between machines, and run across multiple clients. If you've been working with agents locally and wondering what the next step looks like, this is it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 1","logicalValue":"Festival Pavilion, Breakout 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776185877803001f40v","RegistrantKey":"1768245294609build26-1776185877803001f40v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777003136185001uvnL","RegistrantKey":"1768245294609build26-1777003136185001uvnL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK206","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"5ecabed9-25fa-45a5-85e6-44125bfced58","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Patrick Nikoletich, Steve Sanderson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780446984357001HzoY","langLocale":"en-US","sessionId":"1780446984357001HzoY","sessionInstanceId":"1780446984357001HzoY","sessionCode":"BRK206-R1","sessionCodeNormalized":"EN-US-BRK206-R1","title":"Your agent, anywhere: MultiClient, MultiDevice with GitHub Copilot SDK","sortTitle":"your agent, anywhere: multiclient, multidevice with github copilot sdk","description":"Agents are powerful on your machine, but what happens when you need them everywhere else? In this session, we'll show how GitHub Copilot SDK lets you build an agent, embed it in an app, and take it with you across devices and into the cloud. You'll see how to go from a local agent to one you can access on your phone, move between machines, and run across multiple clients. If you've been working with agents locally and wondering what the next step looks like, this is it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The presentation opens with Patrick from the GitHub Copilot CLI and SDK team welcoming the audience and introducing the session’s focus on the Copilot CLI and SDK (<a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>). Patrick outlines that they will discuss how to build and deploy with the SDK, touch on upcoming features, and share follow-up resources. He announces that the GitHub Copilot SDK has reached its General Availability milestone at version 1.0 (<a class=\"medius-aisummary\" data-starttime=\"00:00:32\">00:00:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>), marking the end of a year-long journey from technical preview to official release. The SDK shares the same runtime harness used across various Copilot implementations, such as Copilot CLI and GitHub.com, ensuring a unified experience across tools. Quality and consistency are emphasized as central goals in rolling out SDK improvements and integrating features directly from the CLI.</p>\n\n<p><strong>SDK Capabilities and Adoption:</strong> Patrick elaborates on how the SDK expands beyond GitHub environments to operate across diverse infrastructures, including on-premises systems, containerized environments, multi-tenant services, and even air-gapped government clouds, with or without a GitHub subscription (<a class=\"medius-aisummary\" data-starttime=\"00:02:18\">00:02:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:32\">00:02:32</a>). Initial SDK support covered TypeScript, Python, .NET, and Go, later adding Java and now Rust with the GA release, broadening language versatility (<a class=\"medius-aisummary\" data-starttime=\"00:02:48\">00:02:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>). Patrick notes that collaboration between GitHub and Microsoft engineering teams has widened SDK adoption across major Office products like Excel and Copilot Studio (<a class=\"medius-aisummary\" data-starttime=\"00:03:57\">00:03:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:16\">00:04:16</a>), yielding significant quality improvements, up to 40–50% gains in measured performance for agent-driven coding and non-coding scenarios. This shared “harness” ensures that experiences remain consistent across GitHub and Microsoft’s broader ecosystem, introducing a uniform Copilot feel across different products.</p>\n\n<p><strong>Building and Deploying Agents:</strong> Steve Sanderson takes over the session to demonstrate how developers can use the Copilot SDK to construct multi-agent systems for realistic scenarios (<a class=\"medius-aisummary\" data-starttime=\"00:05:32\">00:05:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>). In his example, Steve simulates a real-estate lead management workflow, illustrating how an AI agent validated inquiries, searched databases, and generated reports. The demonstration highlights how developers can customize system prompts, define agent roles, and introduce custom tools for domain-specific functions such as database searches (<a class=\"medius-aisummary\" data-starttime=\"00:10:35\">00:10:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:02\">00:13:02</a>). Steve shows that agents can run in parallel and that tooling can be tightly controlled for safety. By defining lightweight permissions and embedding only necessary plugins, users can maintain strict boundaries, ensuring that each agent operates autonomously within set constraints without compromising host systems. The SDK’s flexibility allows for easy integration of custom functions and built-in tools, echoing the power of the Copilot CLI runtime.</p>\n\n<p><strong>Sandboxing, Security, and Isolation Techniques:</strong> Steve continues to explore system security in depth, demonstrating isolation strategies such as OS-managed sandboxing through the new MXC cross-platform library (<a class=\"medius-aisummary\" data-starttime=\"00:16:36\">00:16:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:04\">00:18:04</a>). He creates a controlled vulnerability scenario where an AI agent might inadvertently execute malicious instructions, illustrating why sandboxing is essential. The sandbox prevents unauthorized file operations even when an agent’s prompts have been compromised (<a class=\"medius-aisummary\" data-starttime=\"00:18:46\">00:18:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:26\">00:20:26</a>). Steve then presents advanced isolation solutions, including per-session virtualized file systems that allow each AI session to run within its own contained environment without cross-access to host resources (<a class=\"medius-aisummary\" data-starttime=\"00:21:19\">00:21:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>). Participants see how this model bolsters both security and operational resilience, enabling safe, scalable deployments that can migrate or be restored across distributed servers, ideal for multi-user and enterprise-grade applications.</p>\n\n<p><strong>Future Direction and Innovations:</strong> Returning to the stage, Patrick outlines the roadmap for Copilot SDK and GitHub’s Copilot ecosystem (<a class=\"medius-aisummary\" data-starttime=\"00:30:24\">00:30:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:31:02\">00:31:02</a>). He describes an ongoing unification effort across Visual Studio, VS Code, and GitHub.com to ensure that agentic experiences feel consistent everywhere. Future updates will emphasize cloud-based remote agent execution, allowing developers to spawn and monitor isolated sessions across devices and browsers (<a class=\"medius-aisummary\" data-starttime=\"00:32:01\">00:32:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:31\">00:33:31</a>). New investments will also make agents more portable and composable through plugins, memory sharing, and compliance-ready features. Additionally, the team is rewriting the Copilot runtime in Rust to achieve 10x performance gains and memory reductions, improving portability across environments (<a class=\"medius-aisummary\" data-starttime=\"00:35:24\">00:35:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:33\">00:36:33</a>). A lightweight, sandboxed Rust-based runtime will enable running secure Copilot instances even inside browsers or restricted execution contexts, signaling a future of distributed yet cohesive AI tooling within and beyond GitHub’s ecosystem.</p>\n\n<p><strong>Conclusion and Resources:</strong> Patrick closes by summarizing the progress achieved and by providing useful resources for developers, including repository links, best practice guides, and documentation for customizing agents and plugins (<a class=\"medius-aisummary\" data-starttime=\"00:37:44\">00:37:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:38:09\">00:38:09</a>). He underscores that Copilot SDK is designed for openness, adaptability, and sustainable performance across all developer workflows. Attendees are invited to follow the project documentation at GitHub or reach out directly for further discussions. The session wraps with an optimistic outlook toward continued innovation in Copilot SDK, emphasizing security, scalability, and unified coding experiences for future generations of intelligent developer tools.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":3,"TimeSlot":"23:00 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:45:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":true,"onDemand":"https://medius.microsoft.com/Embed/video-nc/77be1034-59af-4388-bdb6-37a99f2acae7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/77be1034-59af-4388-bdb6-37a99f2acae7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/77be1034-59af-4388-bdb6-37a99f2acae7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/77be1034-59af-4388-bdb6-37a99f2acae7","speakerIds":["1768245294609build26-1777003136185001uvnL","1768245294609build26-1776185877803001f40v"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/77be1034-59af-4388-bdb6-37a99f2acae7","ogTitle":"Your agent, anywhere: MultiClient, MultiDevice with GitHub Copilot SDK","ogDescription":"Agents are powerful on your machine, but what happens when you need them everywhere else? In this session, we'll show how GitHub Copilot SDK lets you build an agent, embed it in an app, and take it with you across devices and into the cloud. You'll see how to go from a local agent to one you can access on your phone, move between machines, and run across multiple clients. If you've been working with agents locally and wondering what the next step looks like, this is it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"},{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Breakout","logicalValue":"Breakout"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Breakout 3","logicalValue":"Festival Pavilion, Breakout 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776185877803001f40v","RegistrantKey":"1768245294609build26-1776185877803001f40v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777003136185001uvnL","RegistrantKey":"1768245294609build26-1777003136185001uvnL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/BRK206","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"f8c58625-6e6c-4983-b584-408486f78bee","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mahdi Ghodsi, Eda Zhou","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778175897555001qe39","langLocale":"en-US","sessionId":"1778175897555001qe39","sessionInstanceId":"1778175897555001qe39","sessionCode":"LABSP583-R2","sessionCodeNormalized":"EN-US-LABSP583-R2","title":"Build a custom AI agent with open-weight models and OpenClaw","sortTitle":"build a custom ai agent with open-weight models and openclaw","description":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nNote: This Lab includes breakfast at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175897555001qe39_20260521160952919.avif","speakerIds":["1768245294609build26-1778539204572001y3am","1768245294609build26-1778539327865001THkU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.769Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175897555001qe39_20260521160952919.avif","ogTitle":"Build a custom AI agent with open-weight models and OpenClaw","ogDescription":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nNote: This Lab includes breakfast at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539204572001y3am","RegistrantKey":"1768245294609build26-1778539204572001y3am"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539327865001THkU","RegistrantKey":"1768245294609build26-1778539327865001THkU"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"4d0504df-bcaf-48ff-862d-d0c275e60359","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mahdi Ghodsi, Eda Zhou","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778175955770001w6mO","langLocale":"en-US","sessionId":"1778175955770001w6mO","sessionInstanceId":"1778175955770001w6mO","sessionCode":"LABSP583-R1","sessionCodeNormalized":"EN-US-LABSP583-R1","title":"Build a custom AI agent with open-weight models and OpenClaw","sortTitle":"build a custom ai agent with open-weight models and openclaw","description":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nJoin us for late night bytes!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175955770001w6mO_20260521161005749.avif","speakerIds":["1768245294609build26-1778539204572001y3am","1768245294609build26-1778539327865001THkU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.77Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175955770001w6mO_20260521161005749.avif","ogTitle":"Build a custom AI agent with open-weight models and OpenClaw","ogDescription":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nJoin us for late night bytes!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539327865001THkU","RegistrantKey":"1768245294609build26-1778539327865001THkU"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539204572001y3am","RegistrantKey":"1768245294609build26-1778539204572001y3am"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"c313a048-9fbe-423f-a83a-e5db1e62c030","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mahdi Ghodsi, Eda Zhou","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176105728001AVFb","langLocale":"en-US","sessionId":"1778176105728001AVFb","sessionInstanceId":"1778176105728001AVFb","sessionCode":"LABSP583-R3","sessionCodeNormalized":"EN-US-LABSP583-R3","title":"Build a custom AI agent with open-weight models and OpenClaw","sortTitle":"build a custom ai agent with open-weight models and openclaw","description":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nJoin us for the Bytes at Build reception before the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176105728001AVFb_20260521161019966.avif","speakerIds":["1768245294609build26-1778539204572001y3am","1768245294609build26-1778539327865001THkU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.77Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176105728001AVFb_20260521161019966.avif","ogTitle":"Build a custom AI agent with open-weight models and OpenClaw","ogDescription":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nJoin us for the Bytes at Build reception before the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539204572001y3am","RegistrantKey":"1768245294609build26-1778539204572001y3am"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539327865001THkU","RegistrantKey":"1768245294609build26-1778539327865001THkU"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"310d6e3d-a178-499b-88b5-71ea62dfb8e6","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mahdi Ghodsi, Eda Zhou","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778175457810001loW5","langLocale":"en-US","sessionId":"1778175457810001loW5","sessionInstanceId":"1778175457810001loW5","sessionCode":"LABSP583","sessionCodeNormalized":"EN-US-LABSP583","title":"Build a custom AI agent with open-weight models and OpenClaw","sortTitle":"build a custom ai agent with open-weight models and openclaw","description":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nNote: This Lab includes lunch at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175457810001loW5_20260521160940793.avif","speakerIds":["1768245294609build26-1778539204572001y3am","1768245294609build26-1778539327865001THkU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.769Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778175457810001loW5_20260521160940793.avif","ogTitle":"Build a custom AI agent with open-weight models and OpenClaw","ogDescription":"Build and run your own AI agent by hosting open-weight models on AMD GPU infrastructure. In this hands-on lab, you’ll assemble a tool-using agent, configure model hosting, and connect it to real workflows. Learn how to customize behavior, reduce API costs, and deploy agents with greater privacy and control using OpenClaw.\n\nNote: This Lab includes lunch at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539204572001y3am","RegistrantKey":"1768245294609build26-1778539204572001y3am"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539327865001THkU","RegistrantKey":"1768245294609build26-1778539327865001THkU"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"144b6292-6e64-4455-be0f-22477c8b0aaf","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Matt Calder, Someleze Diko","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774304130264001mWLS","langLocale":"en-US","sessionId":"1774304130264001mWLS","sessionInstanceId":"1774304130264001mWLS","sessionCode":"LAB513","sessionCodeNormalized":"EN-US-LAB513","title":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","sortTitle":"build an ai app with azure sql hyperscale, microsoft fabric, & foundry","description":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/1dcf3ea5-3abd-4d0e-9c21-fc98b10dd73b","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774304130264001mWLS_20260512023515945.avif","speakerIds":["1768245294609build26-1753129132101001Cp9w","1768245294609build26-1762290823994001cHor"],"sessionLinks":[],"relatedSessionCodes":["BRK223"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.636Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774304130264001mWLS_20260512023515945.avif","ogTitle":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","ogDescription":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1753129132101001Cp9w","RegistrantKey":"1768245294609build26-1753129132101001Cp9w"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762290823994001cHor","RegistrantKey":"1768245294609build26-1762290823994001cHor"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"7607897e-ec76-4f46-9f0d-bcbe4a480bbe","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Matt Calder, Someleze Diko","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358396196001XdQD","langLocale":"en-US","sessionId":"1776358396196001XdQD","sessionInstanceId":"1776358396196001XdQD","sessionCode":"LAB513-R2","sessionCodeNormalized":"EN-US-LAB513-R2","title":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","sortTitle":"build an ai app with azure sql hyperscale, microsoft fabric, & foundry","description":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358396196001XdQD_20260512030602856.avif","speakerIds":["1768245294609build26-1753129132101001Cp9w","1768245294609build26-1762290823994001cHor"],"sessionLinks":[],"relatedSessionCodes":["BRK223"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358396196001XdQD_20260512030602856.avif","ogTitle":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","ogDescription":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1753129132101001Cp9w","RegistrantKey":"1768245294609build26-1753129132101001Cp9w"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762290823994001cHor","RegistrantKey":"1768245294609build26-1762290823994001cHor"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"dd02be83-5af7-40a5-b3ac-fc4a38052abc","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Matt Calder, Someleze Diko","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776356807542001Ld5F","langLocale":"en-US","sessionId":"1776356807542001Ld5F","sessionInstanceId":"1776356807542001Ld5F","sessionCode":"LAB513-R1","sessionCodeNormalized":"EN-US-LAB513-R1","title":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","sortTitle":"build an ai app with azure sql hyperscale, microsoft fabric, & foundry","description":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776356807542001Ld5F_20260512030549424.avif","speakerIds":["1768245294609build26-1753129132101001Cp9w","1768245294609build26-1762290823994001cHor"],"sessionLinks":[],"relatedSessionCodes":["BRK223"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776356807542001Ld5F_20260512030549424.avif","ogTitle":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","ogDescription":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1753129132101001Cp9w","RegistrantKey":"1768245294609build26-1753129132101001Cp9w"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762290823994001cHor","RegistrantKey":"1768245294609build26-1762290823994001cHor"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"9606d3b6-9c79-47e7-9a83-68933ba06e63","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Matt Calder, Someleze Diko","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358450469001HOyx","langLocale":"en-US","sessionId":"1776358450469001HOyx","sessionInstanceId":"1776358450469001HOyx","sessionCode":"LAB513-R3","sessionCodeNormalized":"EN-US-LAB513-R3","title":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","sortTitle":"build an ai app with azure sql hyperscale, microsoft fabric, & foundry","description":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358450469001HOyx_20260512030619222.avif","speakerIds":["1768245294609build26-1753129132101001Cp9w","1768245294609build26-1762290823994001cHor"],"sessionLinks":[],"relatedSessionCodes":["BRK223"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358450469001HOyx_20260512030619222.avif","ogTitle":"Build an AI app with Azure SQL Hyperscale, Microsoft Fabric, & Foundry","ogDescription":"Build an AI-powered FAQ assistant using Azure SQL Database vector search and Retrieval-Augmented Generation (RAG) with Azure OpenAI Service. Use GitHub Copilot to speed T-SQL, mirror data into Microsoft Fabric (OneLake) for analytics, apply governance with Microsoft Purview, and orchestrate the workflow with Microsoft Foundry Agents.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1753129132101001Cp9w","RegistrantKey":"1768245294609build26-1753129132101001Cp9w"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762290823994001cHor","RegistrantKey":"1768245294609build26-1762290823994001cHor"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"bc912e0f-adee-4156-b7f7-3a1115684a9e","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Harmke Alkemade, Dimitri Maltezakis, Anton Slutsky","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778168629147001kJ0I","langLocale":"en-US","sessionId":"1778168629147001kJ0I","sessionInstanceId":"1778168629147001kJ0I","sessionCode":"LABSP580","sessionCodeNormalized":"EN-US-LABSP580","title":"Build and deploy reasoning agents with NVIDIA Nemotron and Foundry","sortTitle":"build and deploy reasoning agents with nvidia nemotron and foundry","description":"Build and deploy intelligent AI agents using NVIDIA AI-Q Blueprint and Nemotron with Microsoft Foundry, Azure Container Apps, and Azure AI Search. In this hands-on lab, you'll create an agent for enterprise deep research that combines AI reasoning, multimodal data extraction, and retrieval-augmented generation to synthesize data into actionable insights. Learn how to connect data sources and run inference at scale—then apply these patterns to your own production workflows.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778168629147001kJ0I_20260514205136610.avif","speakerIds":["1768245294609build26-1761310297823001XEmt","1768245294609build26-1758201813158001HGTU","1768245294609build26-1767948275574001iNqE"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.768Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778168629147001kJ0I_20260514205136610.avif","ogTitle":"Build and deploy reasoning agents with NVIDIA Nemotron and Foundry","ogDescription":"Build and deploy intelligent AI agents using NVIDIA AI-Q Blueprint and Nemotron with Microsoft Foundry, Azure Container Apps, and Azure AI Search. In this hands-on lab, you'll create an agent for enterprise deep research that combines AI reasoning, multimodal data extraction, and retrieval-augmented generation to synthesize data into actionable insights. Learn how to connect data sources and run inference at scale—then apply these patterns to your own production workflows.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"NVIDIA @ 308","logicalValue":"NVIDIA @ 308"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758201813158001HGTU","RegistrantKey":"1768245294609build26-1758201813158001HGTU"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1761310297823001XEmt","RegistrantKey":"1768245294609build26-1761310297823001XEmt"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767948275574001iNqE","RegistrantKey":"1768245294609build26-1767948275574001iNqE"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"c868ecea-f6e9-486b-a9fa-07b8bb858b56","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Harmke Alkemade, Dimitri Maltezakis, Anton Slutsky","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778168728369001Vl3U","langLocale":"en-US","sessionId":"1778168728369001Vl3U","sessionInstanceId":"1778168728369001Vl3U","sessionCode":"LABSP580-R1","sessionCodeNormalized":"EN-US-LABSP580-R1","title":"Build and deploy reasoning agents with NVIDIA Nemotron and Foundry","sortTitle":"build and deploy reasoning agents with nvidia nemotron and foundry","description":"Build and deploy intelligent AI agents using NVIDIA AI-Q Blueprint and Nemotron with Microsoft Foundry, Azure Container Apps, and Azure AI Search. In this hands-on lab, you'll create an agent for enterprise deep research that combines AI reasoning, multimodal data extraction, and retrieval-augmented generation to synthesize data into actionable insights. Learn how to connect data sources and run inference at scale—then apply these patterns to your own production workflows.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778168728369001Vl3U_20260514205150715.avif","speakerIds":["1768245294609build26-1761310297823001XEmt","1768245294609build26-1758201813158001HGTU","1768245294609build26-1767948275574001iNqE"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.77Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778168728369001Vl3U_20260514205150715.avif","ogTitle":"Build and deploy reasoning agents with NVIDIA Nemotron and Foundry","ogDescription":"Build and deploy intelligent AI agents using NVIDIA AI-Q Blueprint and Nemotron with Microsoft Foundry, Azure Container Apps, and Azure AI Search. In this hands-on lab, you'll create an agent for enterprise deep research that combines AI reasoning, multimodal data extraction, and retrieval-augmented generation to synthesize data into actionable insights. Learn how to connect data sources and run inference at scale—then apply these patterns to your own production workflows.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"NVIDIA @ 308","logicalValue":"NVIDIA @ 308"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758201813158001HGTU","RegistrantKey":"1768245294609build26-1758201813158001HGTU"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1761310297823001XEmt","RegistrantKey":"1768245294609build26-1761310297823001XEmt"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767948275574001iNqE","RegistrantKey":"1768245294609build26-1767948275574001iNqE"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"fba20aca-7c13-42f9-a7d9-765e8ad55db7","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Satya Devineni, Krishna Sivakumar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176305449001q7MU","langLocale":"en-US","sessionId":"1778176305449001q7MU","sessionInstanceId":"1778176305449001q7MU","sessionCode":"LABSP586-R1","sessionCodeNormalized":"EN-US-LABSP586-R1","title":"Build with local AI agents using OpenCode and Lemonade","sortTitle":"build with local ai agents using opencode and lemonade","description":"Harness the on-device power of the AMD Ryzen™ AI MAX+ processors  to run large coding models locally using Lemonade Server — no cloud, no API keys, all local. Learn how to architect specialized AI agents with OpenCode and watch them collaborate to build something you'll actually want to play. Buckle up - this one's going to be fun.\n\nNote: This includes lunch before the Lab at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176305449001q7MU_20260602111632073.avif","speakerIds":["1768245294609build26-1778539934408001K0UV","1768245294609build26-1778539857024001POJq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.771Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176305449001q7MU_20260602111632073.avif","ogTitle":"Build with local AI agents using OpenCode and Lemonade","ogDescription":"Harness the on-device power of the AMD Ryzen™ AI MAX+ processors  to run large coding models locally using Lemonade Server — no cloud, no API keys, all local. Learn how to architect specialized AI agents with OpenCode and watch them collaborate to build something you'll actually want to play. Buckle up - this one's going to be fun.\n\nNote: This includes lunch before the Lab at Greens Restaurant—an iconic, seasonal, and organic dining experience—plan to arrive early and enjoy.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539857024001POJq","RegistrantKey":"1768245294609build26-1778539857024001POJq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539934408001K0UV","RegistrantKey":"1768245294609build26-1778539934408001K0UV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"f0c6c223-861e-4fd0-9458-1f1b3d25bc8e","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Satya Devineni, Krishna Sivakumar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176421239001iPa5","langLocale":"en-US","sessionId":"1778176421239001iPa5","sessionInstanceId":"1778176421239001iPa5","sessionCode":"LABSP586","sessionCodeNormalized":"EN-US-LABSP586","title":"Build with local AI agents using OpenCode and Lemonade","sortTitle":"build with local ai agents using opencode and lemonade","description":"Harness the on-device power of the AMD Ryzen™ AI MAX+ processors  to run large coding models locally using Lemonade Server — no cloud, no API keys, all local. Learn how to architect specialized AI agents with OpenCode and watch them collaborate to build something you'll actually want to play. Buckle up - this one's going to be fun.\n\nJoin us for late night bytes! \n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176421239001iPa5_20260602111640457.avif","speakerIds":["1768245294609build26-1778539934408001K0UV","1768245294609build26-1778539857024001POJq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.771Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176421239001iPa5_20260602111640457.avif","ogTitle":"Build with local AI agents using OpenCode and Lemonade","ogDescription":"Harness the on-device power of the AMD Ryzen™ AI MAX+ processors  to run large coding models locally using Lemonade Server — no cloud, no API keys, all local. Learn how to architect specialized AI agents with OpenCode and watch them collaborate to build something you'll actually want to play. Buckle up - this one's going to be fun.\n\nJoin us for late night bytes! \n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539857024001POJq","RegistrantKey":"1768245294609build26-1778539857024001POJq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539934408001K0UV","RegistrantKey":"1768245294609build26-1778539934408001K0UV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"60e4ab30-4d48-4f9a-bb1a-34b6e158f825","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Robert Nishi, Steve Thomas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776362098748001Dfgl","langLocale":"en-US","sessionId":"1776362098748001Dfgl","sessionInstanceId":"1776362098748001Dfgl","sessionCode":"LAB550-R3","sessionCodeNormalized":"EN-US-LAB550-R3","title":"Build, deploy, and scale agents with Windows 365","sortTitle":"build, deploy, and scale agents with windows 365","description":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776362098748001Dfgl_20260512031224898.avif","speakerIds":["1768245294609build26-1768599320778001SufH","1768245294609build26-1775060213552001eZMs"],"sessionLinks":[],"relatedSessionCodes":["OD855"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.382Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776362098748001Dfgl_20260512031224898.avif","ogTitle":"Build, deploy, and scale agents with Windows 365","ogDescription":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775060213552001eZMs","RegistrantKey":"1768245294609build26-1775060213552001eZMs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768599320778001SufH","RegistrantKey":"1768245294609build26-1768599320778001SufH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"8aea7155-7c8b-4a47-b537-a138c1247723","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Robert Nishi, Steve Thomas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018719535001roIg","langLocale":"en-US","sessionId":"1774018719535001roIg","sessionInstanceId":"1774018719535001roIg","sessionCode":"LAB550","sessionCodeNormalized":"EN-US-LAB550","title":"Build, deploy, and scale agents with Windows 365","sortTitle":"build, deploy, and scale agents with windows 365","description":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018719535001roIg_20260512023040880.avif","speakerIds":["1768245294609build26-1768599320778001SufH","1768245294609build26-1775060213552001eZMs"],"sessionLinks":[],"relatedSessionCodes":["OD855"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.859Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018719535001roIg_20260512023040880.avif","ogTitle":"Build, deploy, and scale agents with Windows 365","ogDescription":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775060213552001eZMs","RegistrantKey":"1768245294609build26-1775060213552001eZMs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768599320778001SufH","RegistrantKey":"1768245294609build26-1768599320778001SufH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"00245adb-d07d-4a09-bf41-d75ab8f91458","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Robert Nishi, Steve Thomas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776362026778001xCDa","langLocale":"en-US","sessionId":"1776362026778001xCDa","sessionInstanceId":"1776362026778001xCDa","sessionCode":"LAB550-R2","sessionCodeNormalized":"EN-US-LAB550-R2","title":"Build, deploy, and scale agents with Windows 365","sortTitle":"build, deploy, and scale agents with windows 365","description":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776362026778001xCDa_20260512031215681.avif","speakerIds":["1768245294609build26-1768599320778001SufH","1768245294609build26-1775060213552001eZMs"],"sessionLinks":[],"relatedSessionCodes":["OD855"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.381Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776362026778001xCDa_20260512031215681.avif","ogTitle":"Build, deploy, and scale agents with Windows 365","ogDescription":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775060213552001eZMs","RegistrantKey":"1768245294609build26-1775060213552001eZMs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768599320778001SufH","RegistrantKey":"1768245294609build26-1768599320778001SufH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"e919acf6-9eaa-4bf0-9fc0-5af58f138ce4","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Robert Nishi, Steve Thomas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776361968786001ZRmb","langLocale":"en-US","sessionId":"1776361968786001ZRmb","sessionInstanceId":"1776361968786001ZRmb","sessionCode":"LAB550-R1","sessionCodeNormalized":"EN-US-LAB550-R1","title":"Build, deploy, and scale agents with Windows 365","sortTitle":"build, deploy, and scale agents with windows 365","description":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361968786001ZRmb_20260512031202259.avif","speakerIds":["1768245294609build26-1768599320778001SufH","1768245294609build26-1775060213552001eZMs"],"sessionLinks":[],"relatedSessionCodes":["OD855"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.382Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361968786001ZRmb_20260512031202259.avif","ogTitle":"Build, deploy, and scale agents with Windows 365","ogDescription":"Learn how Windows 365 delivers secure and preconfigured computing environments for AI agents, and how you can combine that with computer use to accelerate the creation and deployment of AI agents in enterprise environments. Through technical demos, this session shows how to set up Cloud PCs and how agentic AI interacts with live apps and UIs, while keeping governance, identity, and enterprise controls built in.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775060213552001eZMs","RegistrantKey":"1768245294609build26-1775060213552001eZMs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768599320778001SufH","RegistrantKey":"1768245294609build26-1768599320778001SufH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"816d753f-39df-4484-9654-d7ef84cd2d1d","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonathon Frost, Denzil Ribeiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352569123001PtCu","langLocale":"en-US","sessionId":"1776352569123001PtCu","sessionInstanceId":"1776352569123001PtCu","sessionCode":"LAB511-R3","sessionCodeNormalized":"EN-US-LAB511-R3","title":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","sortTitle":"create advanced postgres-powered agentic apps with azure horizondb","description":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352569123001PtCu_20260512030501707.avif","speakerIds":["1768245294609build26-1765362664415001iCOs","1768245294609build26-1765804946228001vzRn"],"sessionLinks":[],"relatedSessionCodes":["LTG413"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352569123001PtCu_20260512030501707.avif","ogTitle":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","ogDescription":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362664415001iCOs","RegistrantKey":"1768245294609build26-1765362664415001iCOs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765804946228001vzRn","RegistrantKey":"1768245294609build26-1765804946228001vzRn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"a7f15d9b-98e2-40df-95d7-1fbba3cedbec","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonathon Frost, Denzil Ribeiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018706600001rvpY","langLocale":"en-US","sessionId":"1774018706600001rvpY","sessionInstanceId":"1774018706600001rvpY","sessionCode":"LAB511","sessionCodeNormalized":"EN-US-LAB511","title":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","sortTitle":"create advanced postgres-powered agentic apps with azure horizondb","description":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018706600001rvpY_20260512021748861.avif","speakerIds":["1768245294609build26-1765362664415001iCOs","1768245294609build26-1765804946228001vzRn"],"sessionLinks":[],"relatedSessionCodes":["LTG413"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.462Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018706600001rvpY_20260512021748861.avif","ogTitle":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","ogDescription":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362664415001iCOs","RegistrantKey":"1768245294609build26-1765362664415001iCOs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765804946228001vzRn","RegistrantKey":"1768245294609build26-1765804946228001vzRn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"867a8e52-10fd-47c7-a4ce-b77c8b60c54a","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonathon Frost, Denzil Ribeiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352494301001L5z6","langLocale":"en-US","sessionId":"1776352494301001L5z6","sessionInstanceId":"1776352494301001L5z6","sessionCode":"LAB511-R2","sessionCodeNormalized":"EN-US-LAB511-R2","title":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","sortTitle":"create advanced postgres-powered agentic apps with azure horizondb","description":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352494301001L5z6_20260512030452284.avif","speakerIds":["1768245294609build26-1765362664415001iCOs","1768245294609build26-1765804946228001vzRn"],"sessionLinks":[],"relatedSessionCodes":["LTG413"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.892Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352494301001L5z6_20260512030452284.avif","ogTitle":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","ogDescription":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362664415001iCOs","RegistrantKey":"1768245294609build26-1765362664415001iCOs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765804946228001vzRn","RegistrantKey":"1768245294609build26-1765804946228001vzRn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"4520998c-ec84-480a-b073-6d759acb4b44","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonathon Frost, Denzil Ribeiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352246322001HrUv","langLocale":"en-US","sessionId":"1776352246322001HrUv","sessionInstanceId":"1776352246322001HrUv","sessionCode":"LAB511-R1","sessionCodeNormalized":"EN-US-LAB511-R1","title":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","sortTitle":"create advanced postgres-powered agentic apps with azure horizondb","description":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352246322001HrUv_20260512030439791.avif","speakerIds":["1768245294609build26-1765362664415001iCOs","1768245294609build26-1765804946228001vzRn"],"sessionLinks":[],"relatedSessionCodes":["LTG413"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.892Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352246322001HrUv_20260512030439791.avif","ogTitle":"Create advanced Postgres-powered agentic apps with Azure HorizonDB","ogDescription":"In this lab, you’ll build a production-ready AI agent using Azure HorizonDB as a unified data, retrieval, and intelligence layer. Implement RAG, hybrid semantic and keyword search, and agentic workflows without stitching external services. Use DiskANN advanced filtering and in-database AI model management to run LLMs and embeddings directly in the database - no vector stores or glue code. Ideal for developers and architects who want simpler stacks, higher performance, and faster AI delivery.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765362664415001iCOs","RegistrantKey":"1768245294609build26-1765362664415001iCOs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765804946228001vzRn","RegistrantKey":"1768245294609build26-1765804946228001vzRn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"c9acb7a8-c43b-4df2-9031-a5a5ad7bf0fd","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Clement Pakkam Isaac, Shuo-jen Wu, Alexander Zeltov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778169059666001AjrV","langLocale":"en-US","sessionId":"1778169059666001AjrV","sessionInstanceId":"1778169059666001AjrV","sessionCode":"LABSP581-R1","sessionCodeNormalized":"EN-US-LABSP581-R1","title":"Create AI-powered video analytics apps with NVIDIA vision agents","sortTitle":"create ai-powered video analytics apps with nvidia vision agents","description":"Build AI-powered video applications using the NVIDIA Video Search and Summarization (VSS) blueprint. In this hands-on lab, you’ll design vision agents that combine VLMs and LLMs to index, search, and summarize video streams, enabling natural language interaction with visual data at scale.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778169059666001AjrV_20260514205203930.avif","speakerIds":["1768245294609build26-1769795134736001dFDH","1768245294609build26-1770228895156001nEYZ","1768245294609build26-1758648423229001ddL5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.769Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778169059666001AjrV_20260514205203930.avif","ogTitle":"Create AI-powered video analytics apps with NVIDIA vision agents","ogDescription":"Build AI-powered video applications using the NVIDIA Video Search and Summarization (VSS) blueprint. In this hands-on lab, you’ll design vision agents that combine VLMs and LLMs to index, search, and summarize video streams, enabling natural language interaction with visual data at scale.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"NVIDIA @ 308","logicalValue":"NVIDIA @ 308"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769795134736001dFDH","RegistrantKey":"1768245294609build26-1769795134736001dFDH"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758648423229001ddL5","RegistrantKey":"1768245294609build26-1758648423229001ddL5"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770228895156001nEYZ","RegistrantKey":"1768245294609build26-1770228895156001nEYZ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"1dc1cb9c-4804-431e-b960-b2599d4df0b9","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Clement Pakkam Isaac, Shuo-jen Wu, Alexander Zeltov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778169123640001kYEh","langLocale":"en-US","sessionId":"1778169123640001kYEh","sessionInstanceId":"1778169123640001kYEh","sessionCode":"LABSP581","sessionCodeNormalized":"EN-US-LABSP581","title":"Create AI-powered video analytics apps with NVIDIA vision agents","sortTitle":"create ai-powered video analytics apps with nvidia vision agents","description":"Build AI-powered video applications using the NVIDIA Video Search and Summarization (VSS) blueprint. In this hands-on lab, you’ll design vision agents that combine VLMs and LLMs to index, search, and summarize video streams, enabling natural language interaction with visual data at scale.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778169123640001kYEh_20260514205217415.avif","speakerIds":["1768245294609build26-1769795134736001dFDH","1768245294609build26-1770228895156001nEYZ","1768245294609build26-1758648423229001ddL5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.77Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778169123640001kYEh_20260514205217415.avif","ogTitle":"Create AI-powered video analytics apps with NVIDIA vision agents","ogDescription":"Build AI-powered video applications using the NVIDIA Video Search and Summarization (VSS) blueprint. In this hands-on lab, you’ll design vision agents that combine VLMs and LLMs to index, search, and summarize video streams, enabling natural language interaction with visual data at scale.\n\nThis Lab is BYOD. Add this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"NVIDIA @ 308","logicalValue":"NVIDIA @ 308"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769795134736001dFDH","RegistrantKey":"1768245294609build26-1769795134736001dFDH"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758648423229001ddL5","RegistrantKey":"1768245294609build26-1758648423229001ddL5"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770228895156001nEYZ","RegistrantKey":"1768245294609build26-1770228895156001nEYZ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"adcc97aa-0ba6-4839-9517-2570ac7ab2dc","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Anthony Bartolo, Gustavo Cordido","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018702532001rk3p","langLocale":"en-US","sessionId":"1774018702532001rk3p","sessionInstanceId":"1774018702532001rk3p","sessionCode":"LAB530","sessionCodeNormalized":"EN-US-LAB530","title":"Engineering agents that reason, act, and adapt","sortTitle":"engineering agents that reason, act, and adapt","description":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018702532001rk3p_20260601103305976.avif","speakerIds":["1768245294609build26-1756137533747001aI1z","1768245294609build26-1756138101616001VtrM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.458Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018702532001rk3p_20260601103305976.avif","ogTitle":"Engineering agents that reason, act, and adapt","ogDescription":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756138101616001VtrM","RegistrantKey":"1768245294609build26-1756138101616001VtrM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137533747001aI1z","RegistrantKey":"1768245294609build26-1756137533747001aI1z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"8471e720-84c3-4b25-aa0a-ffd8c28f8608","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Anthony Bartolo, Gustavo Cordido","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360325086001HnPa","langLocale":"en-US","sessionId":"1776360325086001HnPa","sessionInstanceId":"1776360325086001HnPa","sessionCode":"LAB530-R1","sessionCodeNormalized":"EN-US-LAB530-R1","title":"Engineering agents that reason, act, and adapt","sortTitle":"engineering agents that reason, act, and adapt","description":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360325086001HnPa_20260601103531211.avif","speakerIds":["1768245294609build26-1756137533747001aI1z","1768245294609build26-1756138101616001VtrM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.897Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360325086001HnPa_20260601103531211.avif","ogTitle":"Engineering agents that reason, act, and adapt","ogDescription":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756138101616001VtrM","RegistrantKey":"1768245294609build26-1756138101616001VtrM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137533747001aI1z","RegistrantKey":"1768245294609build26-1756137533747001aI1z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"a942e02b-a989-412a-bd84-7c11ecf76981","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Anthony Bartolo, Gustavo Cordido","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360405653001aIr8","langLocale":"en-US","sessionId":"1776360405653001aIr8","sessionInstanceId":"1776360405653001aIr8","sessionCode":"LAB530-R3","sessionCodeNormalized":"EN-US-LAB530-R3","title":"Engineering agents that reason, act, and adapt","sortTitle":"engineering agents that reason, act, and adapt","description":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360405653001aIr8_20260601103549750.avif","speakerIds":["1768245294609build26-1756137533747001aI1z","1768245294609build26-1756138101616001VtrM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.381Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360405653001aIr8_20260601103549750.avif","ogTitle":"Engineering agents that reason, act, and adapt","ogDescription":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756138101616001VtrM","RegistrantKey":"1768245294609build26-1756138101616001VtrM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137533747001aI1z","RegistrantKey":"1768245294609build26-1756137533747001aI1z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"c12eecfe-4b82-4140-ba9f-c3276ea2a2da","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Anthony Bartolo, Gustavo Cordido","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360366067001DqfL","langLocale":"en-US","sessionId":"1776360366067001DqfL","sessionInstanceId":"1776360366067001DqfL","sessionCode":"LAB530-R2","sessionCodeNormalized":"EN-US-LAB530-R2","title":"Engineering agents that reason, act, and adapt","sortTitle":"engineering agents that reason, act, and adapt","description":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360366067001DqfL_20260601103540974.avif","speakerIds":["1768245294609build26-1756137533747001aI1z","1768245294609build26-1756138101616001VtrM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.897Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360366067001DqfL_20260601103540974.avif","ogTitle":"Engineering agents that reason, act, and adapt","ogDescription":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756138101616001VtrM","RegistrantKey":"1768245294609build26-1756138101616001VtrM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137533747001aI1z","RegistrantKey":"1768245294609build26-1756137533747001aI1z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"9d87d93d-712f-4ace-93c1-2f4630ee717b","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Pamela Fox","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774979447835001AKSK","langLocale":"en-US","sessionId":"1774979447835001AKSK","sessionInstanceId":"1774979447835001AKSK","sessionCode":"LAB532","sessionCodeNormalized":"EN-US-LAB532","title":"From data to context: Agent‑ready knowledge with Foundry IQ","sortTitle":"from data to context: agent‑ready knowledge with foundry iq","description":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774979447835001AKSK_20260512024557679.avif","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1756137289765001yN8u"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.844Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774979447835001AKSK_20260512024557679.avif","ogTitle":"From data to context: Agent‑ready knowledge with Foundry IQ","ogDescription":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137289765001yN8u","RegistrantKey":"1768245294609build26-1756137289765001yN8u"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"d30d7689-d934-4459-a34e-325bb88cfc7a","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Pamela Fox","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360719583001Dofy","langLocale":"en-US","sessionId":"1776360719583001Dofy","sessionInstanceId":"1776360719583001Dofy","sessionCode":"LAB532-R1","sessionCodeNormalized":"EN-US-LAB532-R1","title":"From data to context: Agent‑ready knowledge with Foundry IQ","sortTitle":"from data to context: agent‑ready knowledge with foundry iq","description":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360719583001Dofy_20260512031046201.avif","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1756137289765001yN8u"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.38Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360719583001Dofy_20260512031046201.avif","ogTitle":"From data to context: Agent‑ready knowledge with Foundry IQ","ogDescription":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137289765001yN8u","RegistrantKey":"1768245294609build26-1756137289765001yN8u"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"6767d3bb-f628-4f1c-a46c-dabfe9edb60b","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Pamela Fox","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780442081081001YH4n","langLocale":"en-US","sessionId":"1780442081081001YH4n","sessionInstanceId":"1780442081081001YH4n","sessionCode":"LAB532-R4","sessionCodeNormalized":"EN-US-LAB532-R4","title":"From data to context: Agent‑ready knowledge with Foundry IQ","sortTitle":"from data to context: agent‑ready knowledge with foundry iq","description":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1780442081081001YH4n_20260603161356939.avif","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1756137289765001yN8u"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1780442081081001YH4n_20260603161356939.avif","ogTitle":"From data to context: Agent‑ready knowledge with Foundry IQ","ogDescription":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137289765001yN8u","RegistrantKey":"1768245294609build26-1756137289765001yN8u"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"57fb6339-354d-4215-925d-e00f89b240ff","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Pamela Fox","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360817264001XOd9","langLocale":"en-US","sessionId":"1776360817264001XOd9","sessionInstanceId":"1776360817264001XOd9","sessionCode":"LAB532-R2","sessionCodeNormalized":"EN-US-LAB532-R2","title":"From data to context: Agent‑ready knowledge with Foundry IQ","sortTitle":"from data to context: agent‑ready knowledge with foundry iq","description":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360817264001XOd9_20260512031054925.avif","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1756137289765001yN8u"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.897Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360817264001XOd9_20260512031054925.avif","ogTitle":"From data to context: Agent‑ready knowledge with Foundry IQ","ogDescription":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137289765001yN8u","RegistrantKey":"1768245294609build26-1756137289765001yN8u"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"80aa138f-7492-4684-b88e-56a2ceb2cd0b","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas, Pamela Fox","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776361002095001Etx9","langLocale":"en-US","sessionId":"1776361002095001Etx9","sessionInstanceId":"1776361002095001Etx9","sessionCode":"LAB532-R3","sessionCodeNormalized":"EN-US-LAB532-R3","title":"From data to context: Agent‑ready knowledge with Foundry IQ","sortTitle":"from data to context: agent‑ready knowledge with foundry iq","description":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361002095001Etx9_20260512031104922.avif","speakerIds":["1768245294609build26-1762250455017001UQP1","1768245294609build26-1756137289765001yN8u"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.38Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361002095001Etx9_20260512031104922.avif","ogTitle":"From data to context: Agent‑ready knowledge with Foundry IQ","ogDescription":"What if your coding agent could reason over enterprise knowledge—across files, data, and work signals—without leaving your developer workflow? In this hands-on lab, you’ll build a reusable knowledge base in Foundry IQ using the Microsoft Foundry portal, expose it as an MCP server, and connect it to GitHub Copilot CLI—no custom RAG plumbing required. Experience how Foundry IQ and Copilot CLI work together to turn enterprise knowledge into live context, right where developers build and debug.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137289765001yN8u","RegistrantKey":"1768245294609build26-1756137289765001yN8u"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"6c1fb585-f9ee-4287-8a5b-d3f8a068c2a9","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yun Jung Choi, Lori Fraleigh","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710899001rSa9","langLocale":"en-US","sessionId":"1774018710899001rSa9","sessionInstanceId":"1774018710899001rSa9","sessionCode":"LAB501","sessionCodeNormalized":"EN-US-LAB501","title":"From zero to deployed on Azure with AI agents","sortTitle":"from zero to deployed on azure with ai agents","description":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018710899001rSa9_20260601103315878.avif","speakerIds":["1768245294609build26-1774284994738001pPWz","1768245294609build26-1779480591564001sFBx"],"sessionLinks":[],"relatedSessionCodes":["BRK207"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.788Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018710899001rSa9_20260601103315878.avif","ogTitle":"From zero to deployed on Azure with AI agents","ogDescription":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994738001pPWz","RegistrantKey":"1768245294609build26-1774284994738001pPWz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779480591564001sFBx","RegistrantKey":"1768245294609build26-1779480591564001sFBx"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"27daf575-612a-4db7-8fa1-74920cb127d0","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yun Jung Choi, Lori Fraleigh","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359074567001kHS1","langLocale":"en-US","sessionId":"1776359074567001kHS1","sessionInstanceId":"1776359074567001kHS1","sessionCode":"LAB501-R3","sessionCodeNormalized":"EN-US-LAB501-R3","title":"From zero to deployed on Azure with AI agents","sortTitle":"from zero to deployed on azure with ai agents","description":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359074567001kHS1_20260601103444226.avif","speakerIds":["1768245294609build26-1774284994738001pPWz","1768245294609build26-1779480591564001sFBx"],"sessionLinks":[],"relatedSessionCodes":["BRK207"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359074567001kHS1_20260601103444226.avif","ogTitle":"From zero to deployed on Azure with AI agents","ogDescription":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994738001pPWz","RegistrantKey":"1768245294609build26-1774284994738001pPWz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779480591564001sFBx","RegistrantKey":"1768245294609build26-1779480591564001sFBx"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"f59ffcb5-f38f-4e71-85bd-78092b85b1ef","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yun Jung Choi, Lori Fraleigh","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358971772001LFiZ","langLocale":"en-US","sessionId":"1776358971772001LFiZ","sessionInstanceId":"1776358971772001LFiZ","sessionCode":"LAB501-R1","sessionCodeNormalized":"EN-US-LAB501-R1","title":"From zero to deployed on Azure with AI agents","sortTitle":"from zero to deployed on azure with ai agents","description":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358971772001LFiZ_20260601103428388.avif","speakerIds":["1768245294609build26-1774284994738001pPWz","1768245294609build26-1779480591564001sFBx"],"sessionLinks":[],"relatedSessionCodes":["BRK207"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358971772001LFiZ_20260601103428388.avif","ogTitle":"From zero to deployed on Azure with AI agents","ogDescription":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994738001pPWz","RegistrantKey":"1768245294609build26-1774284994738001pPWz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779480591564001sFBx","RegistrantKey":"1768245294609build26-1779480591564001sFBx"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"363a5334-1340-4ae3-a194-c61042bbc5fc","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yun Jung Choi, Lori Fraleigh","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359029094001ZbMV","langLocale":"en-US","sessionId":"1776359029094001ZbMV","sessionInstanceId":"1776359029094001ZbMV","sessionCode":"LAB501-R2","sessionCodeNormalized":"EN-US-LAB501-R2","title":"From zero to deployed on Azure with AI agents","sortTitle":"from zero to deployed on azure with ai agents","description":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359029094001ZbMV_20260601103436492.avif","speakerIds":["1768245294609build26-1774284994738001pPWz","1768245294609build26-1779480591564001sFBx"],"sessionLinks":[],"relatedSessionCodes":["BRK207"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359029094001ZbMV_20260601103436492.avif","ogTitle":"From zero to deployed on Azure with AI agents","ogDescription":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994738001pPWz","RegistrantKey":"1768245294609build26-1774284994738001pPWz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779480591564001sFBx","RegistrantKey":"1768245294609build26-1779480591564001sFBx"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"26d9ef97-2c3f-4de1-a746-81a8afb6081c","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176056798001ia4D","langLocale":"en-US","sessionId":"1778176056798001ia4D","sessionInstanceId":"1778176056798001ia4D","sessionCode":"LABSP584-R1","sessionCodeNormalized":"EN-US-LABSP584-R1","title":"Generate video with PyTorch and ComfyUI: Build a GPU pipeline","sortTitle":"generate video with pytorch and comfyui: build a gpu pipeline","description":"Build a video generation pipeline using open-source models and GPU acceleration. In this hands-on lab, you’ll use PyTorch and ComfyUI to generate video clips, configure workflows, and optimize performance. Learn how to run and customize video generation pipelines, reduce compute costs, and apply these techniques to real AI scenarios. Use your skills to participate in a challenge to win prizes.\n\nJoin us for the Bytes at Build reception after the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176056798001ia4D_20260512153456630.avif","speakerIds":["1768245294609build26-1778539546100001CNNb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.771Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176056798001ia4D_20260512153456630.avif","ogTitle":"Generate video with PyTorch and ComfyUI: Build a GPU pipeline","ogDescription":"Build a video generation pipeline using open-source models and GPU acceleration. In this hands-on lab, you’ll use PyTorch and ComfyUI to generate video clips, configure workflows, and optimize performance. Learn how to run and customize video generation pipelines, reduce compute costs, and apply these techniques to real AI scenarios. Use your skills to participate in a challenge to win prizes.\n\nJoin us for the Bytes at Build reception after the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539546100001CNNb","RegistrantKey":"1768245294609build26-1778539546100001CNNb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"a2673b4d-9ae2-46fc-a829-5eea409deb6c","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176006868001y5Tl","langLocale":"en-US","sessionId":"1778176006868001y5Tl","sessionInstanceId":"1778176006868001y5Tl","sessionCode":"LABSP584","sessionCodeNormalized":"EN-US-LABSP584","title":"Generate video with PyTorch and ComfyUI: Build a GPU pipeline","sortTitle":"generate video with pytorch and comfyui: build a gpu pipeline","description":"Build a video generation pipeline using open-source models and GPU acceleration. In this hands-on lab, you’ll use PyTorch and ComfyUI to generate video clips, configure workflows, and optimize performance. Learn how to run and customize video generation pipelines, reduce compute costs, and apply these techniques to real AI scenarios. Use your skills to participate in a challenge to win prizes.\n\nJoin us for the Bytes at Build reception after the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176006868001y5Tl_20260512153444829.avif","speakerIds":["1768245294609build26-1778539546100001CNNb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.77Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176006868001y5Tl_20260512153444829.avif","ogTitle":"Generate video with PyTorch and ComfyUI: Build a GPU pipeline","ogDescription":"Build a video generation pipeline using open-source models and GPU acceleration. In this hands-on lab, you’ll use PyTorch and ComfyUI to generate video clips, configure workflows, and optimize performance. Learn how to run and customize video generation pipelines, reduce compute costs, and apply these techniques to real AI scenarios. Use your skills to participate in a challenge to win prizes.\n\nJoin us for the Bytes at Build reception after the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778539546100001CNNb","RegistrantKey":"1768245294609build26-1778539546100001CNNb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"122b2878-25a7-4fc6-b209-1f0cdc5704d8","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell, Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359765604001WmQt","langLocale":"en-US","sessionId":"1776359765604001WmQt","sessionInstanceId":"1776359765604001WmQt","sessionCode":"LAB520-R2","sessionCodeNormalized":"EN-US-LAB520-R2","title":"Get Started with Models in Microsoft Foundry to Build AI Apps","sortTitle":"get started with models in microsoft foundry to build ai apps","description":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359765604001WmQt_20260512030850171.avif","speakerIds":["1768245294609build26-1772472228939001By6I","1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359765604001WmQt_20260512030850171.avif","ogTitle":"Get Started with Models in Microsoft Foundry to Build AI Apps","ogDescription":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"7a5a4152-2c8f-4c61-8064-d11e92c5152f","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell, Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018714689001r71M","langLocale":"en-US","sessionId":"1774018714689001r71M","sessionInstanceId":"1774018714689001r71M","sessionCode":"LAB520","sessionCodeNormalized":"EN-US-LAB520","title":"Get Started with Models in Microsoft Foundry to Build AI Apps","sortTitle":"get started with models in microsoft foundry to build ai apps","description":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018714689001r71M_20260512022615558.avif","speakerIds":["1768245294609build26-1772472228939001By6I","1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.854Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018714689001r71M_20260512022615558.avif","ogTitle":"Get Started with Models in Microsoft Foundry to Build AI Apps","ogDescription":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"e858155d-7959-4cad-a0f7-e7e6d51e8485","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell, Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359702025001LwJ7","langLocale":"en-US","sessionId":"1776359702025001LwJ7","sessionInstanceId":"1776359702025001LwJ7","sessionCode":"LAB520-R1","sessionCodeNormalized":"EN-US-LAB520-R1","title":"Get Started with Models in Microsoft Foundry to Build AI Apps","sortTitle":"get started with models in microsoft foundry to build ai apps","description":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359702025001LwJ7_20260512030832747.avif","speakerIds":["1768245294609build26-1772472228939001By6I","1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359702025001LwJ7_20260512030832747.avif","ogTitle":"Get Started with Models in Microsoft Foundry to Build AI Apps","ogDescription":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"b4f05817-6f9d-4fdb-b87e-80054b3c9a64","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell, Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359808899001Crgd","langLocale":"en-US","sessionId":"1776359808899001Crgd","sessionInstanceId":"1776359808899001Crgd","sessionCode":"LAB520-R3","sessionCodeNormalized":"EN-US-LAB520-R3","title":"Get Started with Models in Microsoft Foundry to Build AI Apps","sortTitle":"get started with models in microsoft foundry to build ai apps","description":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359808899001Crgd_20260512030904397.avif","speakerIds":["1768245294609build26-1772472228939001By6I","1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359808899001Crgd_20260512030904397.avif","ogTitle":"Get Started with Models in Microsoft Foundry to Build AI Apps","ogDescription":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"03f55a21-ce44-4ad6-9115-be5b8447d41d","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gova Babu, Jamshed Damkewala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352900455001EisK","langLocale":"en-US","sessionId":"1776352900455001EisK","sessionInstanceId":"1776352900455001EisK","sessionCode":"LAB512-R1","sessionCodeNormalized":"EN-US-LAB512-R1","title":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","sortTitle":"hands-on with the latest cobalt vms – from code to cloud deployment","description":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352900455001EisK_20260526104924384.avif","speakerIds":["1768245294609build26-1776877272377001WqB6","1768245294609build26-1774284985889001po0v"],"sessionLinks":[],"relatedSessionCodes":["BRK226"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.892Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352900455001EisK_20260526104924384.avif","ogTitle":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","ogDescription":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Spark","logicalValue":"Spark"}],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284985889001po0v","RegistrantKey":"1768245294609build26-1774284985889001po0v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776877272377001WqB6","RegistrantKey":"1768245294609build26-1776877272377001WqB6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"b847be05-3c5e-4473-b79c-cf8aa9dd7d52","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gova Babu, Jamshed Damkewala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352984417001WH5S","langLocale":"en-US","sessionId":"1776352984417001WH5S","sessionInstanceId":"1776352984417001WH5S","sessionCode":"LAB512-R2","sessionCodeNormalized":"EN-US-LAB512-R2","title":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","sortTitle":"hands-on with the latest cobalt vms – from code to cloud deployment","description":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352984417001WH5S_20260526104932979.avif","speakerIds":["1768245294609build26-1776877272377001WqB6","1768245294609build26-1774284985889001po0v"],"sessionLinks":[],"relatedSessionCodes":["BRK226"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352984417001WH5S_20260526104932979.avif","ogTitle":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","ogDescription":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Spark","logicalValue":"Spark"}],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284985889001po0v","RegistrantKey":"1768245294609build26-1774284985889001po0v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776877272377001WqB6","RegistrantKey":"1768245294609build26-1776877272377001WqB6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"f51b2ba0-67a8-49c0-b47d-bb503b455d5f","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gova Babu, Jamshed Damkewala, Amar Dhamdhere","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018706761001rzGW","langLocale":"en-US","sessionId":"1774018706761001rzGW","sessionInstanceId":"1774018706761001rzGW","sessionCode":"LAB512","sessionCodeNormalized":"EN-US-LAB512","title":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","sortTitle":"hands-on with the latest cobalt vms – from code to cloud deployment","description":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/bddef0a3-6ce9-4d97-953b-931974dd256d","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018706761001rzGW_20260526104826932.avif","speakerIds":["1768245294609build26-1776877272377001WqB6","1768245294609build26-1774284985889001po0v","1768245294609build26-1768580784927001elb0"],"sessionLinks":[],"relatedSessionCodes":["BRK226"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.462Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018706761001rzGW_20260526104826932.avif","ogTitle":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","ogDescription":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Spark","logicalValue":"Spark"}],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284985889001po0v","RegistrantKey":"1768245294609build26-1774284985889001po0v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768580784927001elb0","RegistrantKey":"1768245294609build26-1768580784927001elb0"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776877272377001WqB6","RegistrantKey":"1768245294609build26-1776877272377001WqB6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"06dca713-9a57-46c2-94bd-518c9cd2615a","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gova Babu, Jamshed Damkewala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776353082847001qHtf","langLocale":"en-US","sessionId":"1776353082847001qHtf","sessionInstanceId":"1776353082847001qHtf","sessionCode":"LAB512-R3","sessionCodeNormalized":"EN-US-LAB512-R3","title":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","sortTitle":"hands-on with the latest cobalt vms – from code to cloud deployment","description":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776353082847001qHtf_20260526104941010.avif","speakerIds":["1768245294609build26-1776877272377001WqB6","1768245294609build26-1774284985889001po0v"],"sessionLinks":[],"relatedSessionCodes":["BRK226"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776353082847001qHtf_20260526104941010.avif","ogTitle":"Hands-on with the latest Cobalt VMs – from code to cloud deployment","ogDescription":"Roll up your sleeves and get hands-on with the latest Azure Cobalt VMs. Provision your first Cobalt VM and explore what makes Arm tick in the Cloud. From there you'll build multi-arch container images with Docker, push them to ACR, and deploy to AKS clusters running on Cobalt VMs. Next, wire up a CI/CD pipeline that builds, tests, and deploys to Arm. Then go further—deploy a full-stack web app and serve AI predictions using ONNX Runtime on Arm. Walk out ready to build on Cobalt VMs.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Spark","logicalValue":"Spark"}],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284985889001po0v","RegistrantKey":"1768245294609build26-1774284985889001po0v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776877272377001WqB6","RegistrantKey":"1768245294609build26-1776877272377001WqB6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d44c149c-03ca-4f47-b936-4dcb0e8afa60","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alicia Frame, Chris Noring","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018714998001rxIG","langLocale":"en-US","sessionId":"1774018714998001rxIG","sessionInstanceId":"1774018714998001rxIG","sessionCode":"LAB521","sessionCodeNormalized":"EN-US-LAB521","title":"Improving agent behavior using reinforcement learning from traces","sortTitle":"improving agent behavior using reinforcement learning from traces","description":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018714998001rxIG_20260602111507522.avif","speakerIds":["1768245294609build26-1772039473777001dTM9","1768245294609build26-1756137744474001IoYw"],"sessionLinks":[],"relatedSessionCodes":["BRK231"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.854Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018714998001rxIG_20260602111507522.avif","ogTitle":"Improving agent behavior using reinforcement learning from traces","ogDescription":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137744474001IoYw","RegistrantKey":"1768245294609build26-1756137744474001IoYw"}],"repeatedSessions":[],"relatedResources":[{"link":"https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/fine-tuning?tabs=oai-sdk&pivots=programming-language-studio","backpackLink":null,"label":"Fine Tuning Docs","category":"Document"},{"link":"https://github.com/microsoft-foundry/fine-tuning","backpackLink":null,"label":"Sample code & the agent skill for Fine Tuning","category":"Code sample"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"44eca0a2-7560-4375-8427-f61bc93a241c","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alicia Frame, Chris Noring","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359982094001PTZO","langLocale":"en-US","sessionId":"1776359982094001PTZO","sessionInstanceId":"1776359982094001PTZO","sessionCode":"LAB521-R1","sessionCodeNormalized":"EN-US-LAB521-R1","title":"Improving agent behavior using reinforcement learning from traces","sortTitle":"improving agent behavior using reinforcement learning from traces","description":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359982094001PTZO_20260602111547223.avif","speakerIds":["1768245294609build26-1772039473777001dTM9","1768245294609build26-1756137744474001IoYw"],"sessionLinks":[],"relatedSessionCodes":["BRK231"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359982094001PTZO_20260602111547223.avif","ogTitle":"Improving agent behavior using reinforcement learning from traces","ogDescription":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137744474001IoYw","RegistrantKey":"1768245294609build26-1756137744474001IoYw"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"77eb49d6-1f3e-4296-b788-5ea4a430ffb7","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alicia Frame, Chris Noring","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360092302001Lhai","langLocale":"en-US","sessionId":"1776360092302001Lhai","sessionInstanceId":"1776360092302001Lhai","sessionCode":"LAB521-R3","sessionCodeNormalized":"EN-US-LAB521-R3","title":"Improving agent behavior using reinforcement learning from traces","sortTitle":"improving agent behavior using reinforcement learning from traces","description":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360092302001Lhai_20260602111606636.avif","speakerIds":["1768245294609build26-1772039473777001dTM9","1768245294609build26-1756137744474001IoYw"],"sessionLinks":[],"relatedSessionCodes":["BRK231"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.38Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360092302001Lhai_20260602111606636.avif","ogTitle":"Improving agent behavior using reinforcement learning from traces","ogDescription":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137744474001IoYw","RegistrantKey":"1768245294609build26-1756137744474001IoYw"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"0d47b7c4-9e31-4db5-87dd-db81e249b852","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alicia Frame, Chris Noring","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776360031907001bykn","langLocale":"en-US","sessionId":"1776360031907001bykn","sessionInstanceId":"1776360031907001bykn","sessionCode":"LAB521-R2","sessionCodeNormalized":"EN-US-LAB521-R2","title":"Improving agent behavior using reinforcement learning from traces","sortTitle":"improving agent behavior using reinforcement learning from traces","description":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360031907001bykn_20260602111556459.avif","speakerIds":["1768245294609build26-1772039473777001dTM9","1768245294609build26-1756137744474001IoYw"],"sessionLinks":[],"relatedSessionCodes":["BRK231"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776360031907001bykn_20260602111556459.avif","ogTitle":"Improving agent behavior using reinforcement learning from traces","ogDescription":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772039473777001dTM9","RegistrantKey":"1768245294609build26-1772039473777001dTM9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756137744474001IoYw","RegistrantKey":"1768245294609build26-1756137744474001IoYw"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"fa899a2f-80ea-490e-8939-251c43bd13b2","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Josh Johanning, Tiago Pascoal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711063001rGl0","langLocale":"en-US","sessionId":"1774018711063001rGl0","sessionInstanceId":"1774018711063001rGl0","sessionCode":"LAB502","sessionCodeNormalized":"EN-US-LAB502","title":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","sortTitle":"make github copilot work your way: custom tools, context and workflows","description":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/e4031a6b-20bc-4ce0-98f6-12e78b33064e","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711063001rGl0_20260512022314207.avif","speakerIds":["1768245294609build26-1776108120297001raDy","1768245294609build26-1775753765323001hzD2"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.789Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711063001rGl0_20260512022314207.avif","ogTitle":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","ogDescription":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775753765323001hzD2","RegistrantKey":"1768245294609build26-1775753765323001hzD2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776108120297001raDy","RegistrantKey":"1768245294609build26-1776108120297001raDy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"d243ef6d-7b96-46e1-bf24-dc8e5028988b","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Josh Johanning, Tiago Pascoal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359309310001HGOK","langLocale":"en-US","sessionId":"1776359309310001HGOK","sessionInstanceId":"1776359309310001HGOK","sessionCode":"LAB502-R3","sessionCodeNormalized":"EN-US-LAB502-R3","title":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","sortTitle":"make github copilot work your way: custom tools, context and workflows","description":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359309310001HGOK_20260512030815653.avif","speakerIds":["1768245294609build26-1776108120297001raDy","1768245294609build26-1775753765323001hzD2"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359309310001HGOK_20260512030815653.avif","ogTitle":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","ogDescription":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775753765323001hzD2","RegistrantKey":"1768245294609build26-1775753765323001hzD2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776108120297001raDy","RegistrantKey":"1768245294609build26-1776108120297001raDy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"357b064b-247f-423d-b4e1-30791001b8c2","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Josh Johanning, Tiago Pascoal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359208408001Ltgx","langLocale":"en-US","sessionId":"1776359208408001Ltgx","sessionInstanceId":"1776359208408001Ltgx","sessionCode":"LAB502-R1","sessionCodeNormalized":"EN-US-LAB502-R1","title":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","sortTitle":"make github copilot work your way: custom tools, context and workflows","description":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359208408001Ltgx_20260512030749644.avif","speakerIds":["1768245294609build26-1776108120297001raDy","1768245294609build26-1775753765323001hzD2"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359208408001Ltgx_20260512030749644.avif","ogTitle":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","ogDescription":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Building C, Level 2, Room 205","logicalValue":"Building C, Level 2, Room 205"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775753765323001hzD2","RegistrantKey":"1768245294609build26-1775753765323001hzD2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776108120297001raDy","RegistrantKey":"1768245294609build26-1776108120297001raDy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"1e624fd6-fef2-4336-a129-493f0063fae6","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Josh Johanning, Tiago Pascoal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776359255833001PeKh","langLocale":"en-US","sessionId":"1776359255833001PeKh","sessionInstanceId":"1776359255833001PeKh","sessionCode":"LAB502-R2","sessionCodeNormalized":"EN-US-LAB502-R2","title":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","sortTitle":"make github copilot work your way: custom tools, context and workflows","description":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359255833001PeKh_20260512030802034.avif","speakerIds":["1768245294609build26-1776108120297001raDy","1768245294609build26-1775753765323001hzD2"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776359255833001PeKh_20260512030802034.avif","ogTitle":"Make GitHub Copilot Work Your Way: Custom Tools, Context and Workflows","ogDescription":"GitHub Copilot is powerful out of the box, but custom agents let you shape it to fit your team's exact workflows. In this lab, you'll build agents from scratch: set their roles, write instructions, scope their tools, and plug them into real workflows. You'll create reusable Agent Skills for domain knowledge and connect agents to external services via MCP. Everything works across VS Code, Visual Studio, CLI, and Copilot coding agent. Leave with working agents you can adapt to your own codebase.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775753765323001hzD2","RegistrantKey":"1768245294609build26-1775753765323001hzD2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776108120297001raDy","RegistrantKey":"1768245294609build26-1776108120297001raDy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"2a02cab2-010f-4f09-b740-3074c08a20cd","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nitya Narasimhan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776361357242001qdUU","langLocale":"en-US","sessionId":"1776361357242001qdUU","sessionInstanceId":"1776361357242001qdUU","sessionCode":"LAB540-R1","sessionCodeNormalized":"EN-US-LAB540-R1","title":"Observe, optimize and protect your hosted agents in Microsoft Foundry","sortTitle":"observe, optimize and protect your hosted agents in microsoft foundry","description":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361357242001qdUU_20260522172514783.avif","speakerIds":["1768245294609build26-1756141269010001LleM"],"sessionLinks":[],"relatedSessionCodes":["BRK252"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.381Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361357242001qdUU_20260522172514783.avif","ogTitle":"Observe, optimize and protect your hosted agents in Microsoft Foundry","ogDescription":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"2ead358f-a9c3-4631-8683-c0f0db6d831e","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nitya Narasimhan, Filisha Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776361410272001Daog","langLocale":"en-US","sessionId":"1776361410272001Daog","sessionInstanceId":"1776361410272001Daog","sessionCode":"LAB540-R2","sessionCodeNormalized":"EN-US-LAB540-R2","title":"Observe, optimize and protect your hosted agents in Microsoft Foundry","sortTitle":"observe, optimize and protect your hosted agents in microsoft foundry","description":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361410272001Daog_20260512031131344.avif","speakerIds":["1768245294609build26-1756141269010001LleM","1768245294609build26-1774284995366001pNmt"],"sessionLinks":[],"relatedSessionCodes":["BRK252"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.381Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361410272001Daog_20260512031131344.avif","ogTitle":"Observe, optimize and protect your hosted agents in Microsoft Foundry","ogDescription":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995366001pNmt","RegistrantKey":"1768245294609build26-1774284995366001pNmt"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"d7344f65-fd7e-4a3a-a0b1-c91d5a1790ff","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nitya Narasimhan, Filisha Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717379001rjIM","langLocale":"en-US","sessionId":"1774018717379001rjIM","sessionInstanceId":"1774018717379001rjIM","sessionCode":"LAB540","sessionCodeNormalized":"EN-US-LAB540","title":"Observe, optimize and protect your hosted agents in Microsoft Foundry","sortTitle":"observe, optimize and protect your hosted agents in microsoft foundry","description":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"19:30 - 20:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:45:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/1ae9cf88-892f-437e-a640-0b5671ad39b3","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717379001rjIM_20260512022838442.avif","speakerIds":["1768245294609build26-1756141269010001LleM","1768245294609build26-1774284995366001pNmt"],"sessionLinks":[],"relatedSessionCodes":["BRK252"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.856Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717379001rjIM_20260512022838442.avif","ogTitle":"Observe, optimize and protect your hosted agents in Microsoft Foundry","ogDescription":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995366001pNmt","RegistrantKey":"1768245294609build26-1774284995366001pNmt"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"118c9e87-bb78-4195-bf7c-23be607216cb","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nitya Narasimhan, Filisha Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776361505443001EPkN","langLocale":"en-US","sessionId":"1776361505443001EPkN","sessionInstanceId":"1776361505443001EPkN","sessionCode":"LAB540-R3","sessionCodeNormalized":"EN-US-LAB540-R3","title":"Observe, optimize and protect your hosted agents in Microsoft Foundry","sortTitle":"observe, optimize and protect your hosted agents in microsoft foundry","description":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361505443001EPkN_20260512031149057.avif","speakerIds":["1768245294609build26-1756141269010001LleM","1768245294609build26-1774284995366001pNmt"],"sessionLinks":[],"relatedSessionCodes":["BRK252"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.381Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776361505443001EPkN_20260512031149057.avif","ogTitle":"Observe, optimize and protect your hosted agents in Microsoft Foundry","ogDescription":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995366001pNmt","RegistrantKey":"1768245294609build26-1774284995366001pNmt"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"303e4fee-9942-48fb-8d1f-747eee9b1aca","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Han, Michael Han","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176521688001oio0","langLocale":"en-US","sessionId":"1778176521688001oio0","sessionInstanceId":"1778176521688001oio0","sessionCode":"LABSP585-R1","sessionCodeNormalized":"EN-US-LABSP585-R1","title":"Reinforcement learning and local models on AMD Ryzen™ AI MAX+ with Unsloth","sortTitle":"reinforcement learning and local models on amd ryzen™ ai max+ with unsloth","description":"Learn how to fine-tune AI models fully locally on AMD Ryzen™ AI MAX+ processors using Unsloth in this beginner to intermediate, instructor-led, hands-on workshop with no cloud required. Starting from dataset preparation and prompt formatting, you'll walk through the complete finetuning pipeline step by step. Along the way, you'll gain practical tips from industry experts and wrap up with a short challenge to put your freshly fine-tuned model to the test, all running on your own machine.\n\nNote: This includes lunch after the Lab at Greens Restaurant—an iconic, seasonal, and organic dining experience.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176521688001oio0_20260529232611156.avif","speakerIds":["1768245294609build26-1775582336728001AON2","1768245294609build26-1776323049506001IUbh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.771Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176521688001oio0_20260529232611156.avif","ogTitle":"Reinforcement learning and local models on AMD Ryzen™ AI MAX+ with Unsloth","ogDescription":"Learn how to fine-tune AI models fully locally on AMD Ryzen™ AI MAX+ processors using Unsloth in this beginner to intermediate, instructor-led, hands-on workshop with no cloud required. Starting from dataset preparation and prompt formatting, you'll walk through the complete finetuning pipeline step by step. Along the way, you'll gain practical tips from industry experts and wrap up with a short challenge to put your freshly fine-tuned model to the test, all running on your own machine.\n\nNote: This includes lunch after the Lab at Greens Restaurant—an iconic, seasonal, and organic dining experience.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775582336728001AON2","RegistrantKey":"1768245294609build26-1775582336728001AON2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776323049506001IUbh","RegistrantKey":"1768245294609build26-1776323049506001IUbh"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"a0ec7fdc-6370-40f0-b72d-53a90ef15d1a","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Han, Michael Han","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778176155158001xGtI","langLocale":"en-US","sessionId":"1778176155158001xGtI","sessionInstanceId":"1778176155158001xGtI","sessionCode":"LABSP585","sessionCodeNormalized":"EN-US-LABSP585","title":"Reinforcement learning and local models on AMD Ryzen™ AI MAX+ with Unsloth","sortTitle":"reinforcement learning and local models on amd ryzen™ ai max+ with unsloth","description":"Learn how to fine-tune AI models fully locally on AMD Ryzen™ AI MAX+ processors using Unsloth in this beginner to intermediate, instructor-led, hands-on workshop with no cloud required. Starting from dataset preparation and prompt formatting, you'll walk through the complete finetuning pipeline step by step. Along the way, you'll gain practical tips from industry experts and wrap up with a short challenge to put your freshly fine-tuned model to the test, all running on your own machine.\n\nJoin us for the Bytes at Build reception before the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176155158001xGtI_20260529232542670.avif","speakerIds":["1768245294609build26-1775582336728001AON2","1768245294609build26-1776323049506001IUbh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.771Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778176155158001xGtI_20260529232542670.avif","ogTitle":"Reinforcement learning and local models on AMD Ryzen™ AI MAX+ with Unsloth","ogDescription":"Learn how to fine-tune AI models fully locally on AMD Ryzen™ AI MAX+ processors using Unsloth in this beginner to intermediate, instructor-led, hands-on workshop with no cloud required. Starting from dataset preparation and prompt formatting, you'll walk through the complete finetuning pipeline step by step. Along the way, you'll gain practical tips from industry experts and wrap up with a short challenge to put your freshly fine-tuned model to the test, all running on your own machine.\n\nJoin us for the Bytes at Build reception before the Lab!\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"AMD @ Greens","logicalValue":"AMD @ Greens"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775582336728001AON2","RegistrantKey":"1768245294609build26-1775582336728001AON2"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776323049506001IUbh","RegistrantKey":"1768245294609build26-1776323049506001IUbh"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"a50df6d9-4264-478c-b2e4-073229ae7036","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Pratyush Rawat","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777582642663001tz4F","langLocale":"en-US","sessionId":"1777582642663001tz4F","sessionInstanceId":"1777582642663001tz4F","sessionCode":"LAB514-R4","sessionCodeNormalized":"EN-US-LAB514-R4","title":"Ship AI apps fast with Rayfin, a new managed backend as a service","sortTitle":"ship ai apps fast with rayfin, a new managed backend as a service","description":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1780454779879001kZH6_20260603161415148.avif","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1777584842315001x6I9"],"sessionLinks":[],"relatedSessionCodes":["BRK225"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.389Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1780454779879001kZH6_20260603161415148.avif","ogTitle":"Ship AI apps fast with Rayfin, a new managed backend as a service","ogDescription":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777584842315001x6I9","RegistrantKey":"1768245294609build26-1777584842315001x6I9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"b1ef3b94-462b-40a7-b683-3f47e614fb4a","startDateTime":"2026-06-03T03:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Pratyush Rawat","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777582507758001uo3n","langLocale":"en-US","sessionId":"1777582507758001uo3n","sessionInstanceId":"1777582507758001uo3n","sessionCode":"LAB514-R1","sessionCodeNormalized":"EN-US-LAB514-R1","title":"Ship AI apps fast with Rayfin, a new managed backend as a service","sortTitle":"ship ai apps fast with rayfin, a new managed backend as a service","description":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"03:15 - 04:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T04:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777582507758001uo3n_20260603024910297.avif","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1777584842315001x6I9"],"sessionLinks":[],"relatedSessionCodes":["BRK225"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.39Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777582507758001uo3n_20260603024910297.avif","ogTitle":"Ship AI apps fast with Rayfin, a new managed backend as a service","ogDescription":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777584842315001x6I9","RegistrantKey":"1768245294609build26-1777584842315001x6I9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"97a5b6e1-5477-4839-90e9-66cde9b274b6","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Pratyush Rawat","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777471719673001KaB5","langLocale":"en-US","sessionId":"1777471719673001KaB5","sessionInstanceId":"1777471719673001KaB5","sessionCode":"LAB514","sessionCodeNormalized":"EN-US-LAB514","title":"Ship AI apps fast with Rayfin, a new managed backend as a service","sortTitle":"ship ai apps fast with rayfin, a new managed backend as a service","description":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/5bb6f77b-c87b-4d16-8058-9ecd2d4e3b73","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777471719673001KaB5_20260603024902953.avif","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1777584842315001x6I9"],"sessionLinks":[],"relatedSessionCodes":["BRK225"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.389Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777471719673001KaB5_20260603024902953.avif","ogTitle":"Ship AI apps fast with Rayfin, a new managed backend as a service","ogDescription":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 1","logicalValue":"Gateway Pavilion, Level 1, Lab 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777584842315001x6I9","RegistrantKey":"1768245294609build26-1777584842315001x6I9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"5bf6ebb9-4e82-4413-b58b-d0a817a6b57e","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Pratyush Rawat","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780454779879001kZH6","langLocale":"en-US","sessionId":"1780454779879001kZH6","sessionInstanceId":"1780454779879001kZH6","sessionCode":"LAB514-R3","sessionCodeNormalized":"EN-US-LAB514-R3","title":"Ship AI apps fast with Rayfin, a new managed backend as a service","sortTitle":"ship ai apps fast with rayfin, a new managed backend as a service","description":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777471719673001KaB5_20260603024902953.avif","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1777584842315001x6I9"],"sessionLinks":[],"relatedSessionCodes":["BRK225"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.465Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777471719673001KaB5_20260603024902953.avif","ogTitle":"Ship AI apps fast with Rayfin, a new managed backend as a service","ogDescription":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777584842315001x6I9","RegistrantKey":"1768245294609build26-1777584842315001x6I9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"c84bb8e3-ccec-45ae-bc61-3f3ad3bc1b18","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Pratyush Rawat","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777582571352001yzY4","langLocale":"en-US","sessionId":"1777582571352001yzY4","sessionInstanceId":"1777582571352001yzY4","sessionCode":"LAB514-R2","sessionCodeNormalized":"EN-US-LAB514-R2","title":"Ship AI apps fast with Rayfin, a new managed backend as a service","sortTitle":"ship ai apps fast with rayfin, a new managed backend as a service","description":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777582571352001yzY4_20260603024918113.avif","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1777584842315001x6I9"],"sessionLinks":[],"relatedSessionCodes":["BRK225"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.39Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777582571352001yzY4_20260603024918113.avif","ogTitle":"Ship AI apps fast with Rayfin, a new managed backend as a service","ogDescription":"Build a field technician productivity app with Rayfin using coding agents. Generate a full-stack app on Microsoft Fabric without manually configuring backend services, then connect to OneLake to surface service insights in the experience. Leave with a working app and a repeatable pattern to ship data-driven, production-ready apps faster, without backend complexity.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 2","logicalValue":"Gateway Pavilion, Level 1, Lab 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777584842315001x6I9","RegistrantKey":"1768245294609build26-1777584842315001x6I9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"58c6c4e1-556f-4544-8424-211f96c1ce94","startDateTime":"2026-06-03T16:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sertaç Özercan, Paul Yu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776351873525001qXtv","langLocale":"en-US","sessionId":"1776351873525001qXtv","sessionInstanceId":"1776351873525001qXtv","sessionCode":"LAB510-R1","sessionCodeNormalized":"EN-US-LAB510-R1","title":"Take LLMs from prototype to production on AKS","sortTitle":"take llms from prototype to production on aks","description":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"16:00 - 17:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776351873525001qXtv_20260512030406408.avif","speakerIds":["1768245294609build26-1776333707806001tHSv","1768245294609build26-1774284994279001pbAB"],"sessionLinks":[],"relatedSessionCodes":["BRK222"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.892Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776351873525001qXtv_20260512030406408.avif","ogTitle":"Take LLMs from prototype to production on AKS","ogDescription":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Firehouse","logicalValue":"Firehouse"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776333707806001tHSv","RegistrantKey":"1768245294609build26-1776333707806001tHSv"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994279001pbAB","RegistrantKey":"1768245294609build26-1774284994279001pbAB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"}]},{"@search.score":1.0,"documentId":"57e9a40c-9fad-44ae-8796-444b659e0525","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sertaç Özercan, Paul Yu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776351995050001EpMc","langLocale":"en-US","sessionId":"1776351995050001EpMc","sessionInstanceId":"1776351995050001EpMc","sessionCode":"LAB510-R2","sessionCodeNormalized":"EN-US-LAB510-R2","title":"Take LLMs from prototype to production on AKS","sortTitle":"take llms from prototype to production on aks","description":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"20:15 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776351995050001EpMc_20260512030420453.avif","speakerIds":["1768245294609build26-1776333707806001tHSv","1768245294609build26-1774284994279001pbAB"],"sessionLinks":[],"relatedSessionCodes":["BRK222"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.891Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776351995050001EpMc_20260512030420453.avif","ogTitle":"Take LLMs from prototype to production on AKS","ogDescription":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776333707806001tHSv","RegistrantKey":"1768245294609build26-1776333707806001tHSv"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994279001pbAB","RegistrantKey":"1768245294609build26-1774284994279001pbAB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"}]},{"@search.score":1.0,"documentId":"f2f0377d-8299-4eb2-b12a-f90b0e5a4c48","startDateTime":"2026-06-03T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sertaç Özercan, Paul Yu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776352166471001LgaH","langLocale":"en-US","sessionId":"1776352166471001LgaH","sessionInstanceId":"1776352166471001LgaH","sessionCode":"LAB510-R3","sessionCodeNormalized":"EN-US-LAB510-R3","title":"Take LLMs from prototype to production on AKS","sortTitle":"take llms from prototype to production on aks","description":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352166471001LgaH_20260512030429459.avif","speakerIds":["1768245294609build26-1776333707806001tHSv","1768245294609build26-1774284994279001pbAB"],"sessionLinks":[],"relatedSessionCodes":["BRK222"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.891Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776352166471001LgaH_20260512030429459.avif","ogTitle":"Take LLMs from prototype to production on AKS","ogDescription":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 4","logicalValue":"Gateway Pavilion, Level 1, Lab 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776333707806001tHSv","RegistrantKey":"1768245294609build26-1776333707806001tHSv"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994279001pbAB","RegistrantKey":"1768245294609build26-1774284994279001pbAB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"}]},{"@search.score":1.0,"documentId":"c6458193-2c01-485e-a917-2b20f94e816b","startDateTime":"2026-06-02T23:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sertaç Özercan, Paul Yu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774017128575001xSbY","langLocale":"en-US","sessionId":"1774017128575001xSbY","sessionInstanceId":"1774017128575001xSbY","sessionCode":"LAB510","sessionCodeNormalized":"EN-US-LAB510","title":"Take LLMs from prototype to production on AKS","sortTitle":"take llms from prototype to production on aks","description":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"23:45 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774017128575001xSbY_20260512021100624.avif","speakerIds":["1768245294609build26-1776333707806001tHSv","1768245294609build26-1774284994279001pbAB"],"sessionLinks":[],"relatedSessionCodes":["BRK222"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.454Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774017128575001xSbY_20260512021100624.avif","ogTitle":"Take LLMs from prototype to production on AKS","ogDescription":"Moving an AI model from experiment to production is hard. Learn about AI Runway, an open-source accelerator that simplifies deploying LLMs on Azure Kubernetes Service (AKS). By treating models as native Kubernetes resources, AI Runway offers a single interface that adapts to multiple inference backends. You’ll deploy a production LLM on AKS, implement custom resources for scaling and networking, configure GPU and latency monitoring, and integrate it into CI/CD pipelines.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"sgs","logicalValue":"sgs"}],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776333707806001tHSv","RegistrantKey":"1768245294609build26-1776333707806001tHSv"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994279001pbAB","RegistrantKey":"1768245294609build26-1774284994279001pbAB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"}]},{"@search.score":1.0,"documentId":"3eecb7c0-6443-4bd3-9562-9ddd54e4f449","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christopher Harrison, Toby Padilla","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358909696001FtYf","langLocale":"en-US","sessionId":"1776358909696001FtYf","sessionInstanceId":"1776358909696001FtYf","sessionCode":"LAB500-R3","sessionCodeNormalized":"EN-US-LAB500-R3","title":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","sortTitle":"terminal to party: live-coding with github copilot cli and hydra","description":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358909696001FtYf_20260512030700140.avif","speakerIds":["1768245294609build26-1767984616894001huFl","1768245294609build26-1776124361982001jjff"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358909696001FtYf_20260512030700140.avif","ogTitle":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","ogDescription":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767984616894001huFl","RegistrantKey":"1768245294609build26-1767984616894001huFl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776124361982001jjff","RegistrantKey":"1768245294609build26-1776124361982001jjff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"e79aae3a-eb02-40e3-9b2b-ff36cfbcb6a5","startDateTime":"2026-06-03T01:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christopher Harrison, Toby Padilla","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358774912001CUxy","langLocale":"en-US","sessionId":"1776358774912001CUxy","sessionInstanceId":"1776358774912001CUxy","sessionCode":"LAB500-R1","sessionCodeNormalized":"EN-US-LAB500-R1","title":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","sortTitle":"terminal to party: live-coding with github copilot cli and hydra","description":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"01:30 - 02:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T02:45:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358774912001CUxy_20260512030631915.avif","speakerIds":["1768245294609build26-1767984616894001huFl","1768245294609build26-1776124361982001jjff"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358774912001CUxy_20260512030631915.avif","ogTitle":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","ogDescription":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767984616894001huFl","RegistrantKey":"1768245294609build26-1767984616894001huFl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776124361982001jjff","RegistrantKey":"1768245294609build26-1776124361982001jjff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"16d5971b-1a37-4378-b573-42c653bad799","startDateTime":"2026-06-02T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christopher Harrison, Toby Padilla","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710738001rCVC","langLocale":"en-US","sessionId":"1774018710738001rCVC","sessionInstanceId":"1774018710738001rCVC","sessionCode":"LAB500","sessionCodeNormalized":"EN-US-LAB500","title":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","sortTitle":"terminal to party: live-coding with github copilot cli and hydra","description":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"22:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:15:00Z","durationInMinutes":75,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/8be7327b-86e2-4820-ac70-54b9b8f0056d","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018710738001rCVC_20260512022246166.avif","speakerIds":["1768245294609build26-1767984616894001huFl","1768245294609build26-1776124361982001jjff"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.788Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018710738001rCVC_20260512022246166.avif","ogTitle":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","ogDescription":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767984616894001huFl","RegistrantKey":"1768245294609build26-1767984616894001huFl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776124361982001jjff","RegistrantKey":"1768245294609build26-1776124361982001jjff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"61a212e1-b0d5-4fb4-8610-4197bd8b0bc5","startDateTime":"2026-06-03T17:45:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christopher Harrison, Toby Padilla","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776358861989001xLtg","langLocale":"en-US","sessionId":"1776358861989001xLtg","sessionInstanceId":"1776358861989001xLtg","sessionCode":"LAB500-R2","sessionCodeNormalized":"EN-US-LAB500-R2","title":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","sortTitle":"terminal to party: live-coding with github copilot cli and hydra","description":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":4,"TimeSlot":"17:45 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":75,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358861989001xLtg_20260512030647471.avif","speakerIds":["1768245294609build26-1767984616894001huFl","1768245294609build26-1776124361982001jjff"],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.229Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776358861989001xLtg_20260512030647471.avif","ogTitle":"Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","ogDescription":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nAdd this session to your schedule to reserve your seat, as capacity is limited. Please arrive at least 5 minutes early; unclaimed seats may be released to standby attendees.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lab","logicalValue":"Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Lab 3","logicalValue":"Gateway Pavilion, Level 1, Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767984616894001huFl","RegistrantKey":"1768245294609build26-1767984616894001huFl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776124361982001jjff","RegistrantKey":"1768245294609build26-1776124361982001jjff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"a264e008-8961-4dcc-996c-0a09c3393a93","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Shreenidhi Bindinganavile Ramanuja, Matt Erni","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774041947991001CXQy","langLocale":"en-US","sessionId":"1774041947991001CXQy","sessionInstanceId":"1774041947991001CXQy","sessionCode":"DEM351","sessionCodeNormalized":"EN-US-DEM351","title":"AI Skills Navigator: Accelerate tech skills with personalized learning","sortTitle":"ai skills navigator: accelerate tech skills with personalized learning","description":"Learn more about how AI Skills Navigator integrates with Copilot and supports continuous learning at scale to help you stay ahead of rapidly evolving AI capabilities. Join us to hear how AI is transforming skill building for developers by making it more personalized, efficient, and impactful than ever. Learn based on your preferences including short sessions, podcast-style summaries, or alongside your team. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with Matt Ernie, Senior Technical Program Manager at Microsoft Global Skilling, welcoming viewers and setting the stage for a discussion about transforming how people work with AI (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a>). He contrasts last year’s focus on workforce transformation with today’s emphasis on work transformation, highlighting the need to redesign workflows so AI can deliver measurable outcomes. Matt introduces the <em>AI Skills Navigator</em>—Microsoft’s unified upskilling destination that brings together Microsoft Learn modules, LinkedIn Learning paths, trusted YouTube content, and skilling credentials (<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>). He explains that this resource, integrated with the Learning Agent in Microsoft 365 Copilot, supports contextual learning in the flow of work. To demonstrate its value, Matt introduces his co-presenter, Sri, who will lead a scenario-based walkthrough.</p>\n\n<p><strong>Demonstration Setup and Personalization:</strong> Sri assumes the role of an engineering lead at a company called Zava, responsible for upskilling his team to deliver a production-grade AI agent (<a class=\"medius-aisummary\" data-starttime=\"00:01:35\">00:01:35</a>). He begins from the AI Skills Navigator homepage, showing how users can browse without signing in, with content automatically categorized by role and skill level (<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>). Upon signing in, Sri describes the light onboarding experience, which includes selecting roles (e.g., AI Engineer, Data Scientist) and learning preferences such as “visual,” “hands-on,” or “interactive” styles (<a class=\"medius-aisummary\" data-starttime=\"00:03:22\">00:03:22</a>). These choices personalize the “For You” page to deliver targeted resources—like generating tests in AI or evaluating generative AI models—ensuring learners receive relevant content tailored to their professional goals (<a class=\"medius-aisummary\" data-starttime=\"00:04:08\">00:04:08</a>). He also demonstrates the “Explore Content” section, featuring curated collections organized by levels, expert contributors, and multiple modalities including Microsoft Learn paths, credentials, and videos (<a class=\"medius-aisummary\" data-starttime=\"00:04:50\">00:04:50</a>).</p>\n\n<p><strong>Team Skilling Sessions and Playlist Creation:</strong> Moving to collaborative learning, Sri introduces the <em>Team Skilling</em> feature, which allows managers to curate playlists for their teams with the assistance of an AI agent (<a class=\"medius-aisummary\" data-starttime=\"00:05:42\">00:05:42</a>). By describing his objective—to upskill developers on building agents using Microsoft 365 Copilot, Microsoft Graph, and SharePoint connectors—Sri demonstrates how the system automatically generates a structured 4-week plan with about 8 hours of learning per week (<a class=\"medius-aisummary\" data-starttime=\"00:06:22\">00:06:22</a>). The AI curates modules into sections covering agent design, Graph grounding, evaluation, and telemetry concepts. Once finalized, the playlist is shared with Matt, simulating a collaborative team scenario (<a class=\"medius-aisummary\" data-starttime=\"00:07:18\">00:07:18</a>). This feature showcases how AI can accelerate structured curriculum creation aligned with organizational goals.</p>\n\n<p><strong>Learning Experience and Interactive Coaching:</strong> As the assigned developer, Matt opens the shared playlist within Microsoft Teams and explores multiple modes of learning (<a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>). He accesses a Learn module on Copilot Studio agent building but opts to consume it as a podcast-style audio summary—a dynamic output generated from text-based modules (<a class=\"medius-aisummary\" data-starttime=\"00:09:06\">00:09:06</a>). He also highlights curated YouTube sessions and hybrid skilling courses that blend virtual training with guided discussion via an AI coach (<a class=\"medius-aisummary\" data-starttime=\"00:11:07\">00:11:07</a>). The course assistant provides nudges, recap suggestions, and knowledge checks, helping learners assess progress. Matt demonstrates the coach’s contextual awareness in answering questions relevant to the learner’s position in the curriculum. This interactivity illustrates how AI-supported instructional design simulates real-time mentorship (<a class=\"medius-aisummary\" data-starttime=\"00:12:34\">00:12:34</a>).</p>\n\n<p><strong>AI Integration in Microsoft 365 and Credentials:</strong> Matt and Sri then discuss the newly released <em>Learning Agent in Microsoft 365 Copilot</em>, designed to bring learning directly into the work environment (<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>). The agent leverages “Work IQ” data and personal profiles to recommend activities, resources, and tasks, integrating sources like LinkedIn Learning and enterprise content. It supports on-demand learning requests—such as finding applied skills or AI prompting material—tailored to each user’s workflow (<a class=\"medius-aisummary\" data-starttime=\"00:15:25\">00:15:25</a>). The demonstration continues with the applied skills assessments, which validate practical expertise through hands-on labs (<a class=\"medius-aisummary\" data-starttime=\"00:17:02\">00:17:02</a>), transitioning naturally into Microsoft credentials available within AI Skills Navigator, from “Copilot and Agent Administration Fundamentals” to “Azure AI Apps and Agents Developer Associate” (<a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>). Matt emphasizes how such credentials convert learning into recognized achievements that showcase verified competency.</p>\n\n<p><strong>Progress Tracking and Conclusion:</strong> To close the session, Sri revisits the team leader view that displays combined progress across assigned playlists (<a class=\"medius-aisummary\" data-starttime=\"00:19:16\">00:19:16</a>). He demonstrates how the dashboard allows leaders to identify team members’ progress, detect blockers, and adapt learning strategies across multi-week sprints (<a class=\"medius-aisummary\" data-starttime=\"00:20:04\">00:20:04</a>). The presenters summarize the video, reiterating that AI Skills Navigator provides Microsoft’s central hub for applied skilling—featuring content from Microsoft Learn, certifications, skilling sessions, and AI-assisted playlist generation (<a class=\"medius-aisummary\" data-starttime=\"00:21:13\">00:21:13</a>). They encourage viewers to register, personalize their profiles, and join the upcoming “Skills Fest” event taking place June 8–12 as a next step in advancing their AI learning journey (<a class=\"medius-aisummary\" data-starttime=\"00:22:21\">00:22:21</a>). The session concludes with appreciation for participants and a call to action to begin leveraging AI-driven skilling in daily workflows.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/9dbfe949-af78-47ce-b6ed-97552be1217b","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9dbfe949-af78-47ce-b6ed-97552be1217b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9dbfe949-af78-47ce-b6ed-97552be1217b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9dbfe949-af78-47ce-b6ed-97552be1217b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9dbfe949-af78-47ce-b6ed-97552be1217b","speakerIds":["1768245294609build26-1775501322352001aycE","1768245294609build26-1776095049777001mkE5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.232Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9dbfe949-af78-47ce-b6ed-97552be1217b","ogTitle":"AI Skills Navigator: Accelerate tech skills with personalized learning","ogDescription":"Learn more about how AI Skills Navigator integrates with Copilot and supports continuous learning at scale to help you stay ahead of rapidly evolving AI capabilities. Join us to hear how AI is transforming skill building for developers by making it more personalized, efficient, and impactful than ever. Learn based on your preferences including short sessions, podcast-style summaries, or alongside your team. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775501322352001aycE","RegistrantKey":"1768245294609build26-1775501322352001aycE"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776095049777001mkE5","RegistrantKey":"1768245294609build26-1776095049777001mkE5"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM351","backpackLink":null,"label":"Recommended next step","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Personalization","logicalValue":"Personalization"}]},{"@search.score":1.0,"documentId":"81ed346f-07ec-4cbc-ba3d-20be3a08569e","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nagkumar Arkalgud, Hanchi Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717215001rOiT","langLocale":"en-US","sessionId":"1774018717215001rOiT","sessionInstanceId":"1774018717215001rOiT","sessionCode":"DEM341","sessionCodeNormalized":"EN-US-DEM341","title":"Any agent, any cloud: Standardized tracing with Foundry+OpenTelemetry","sortTitle":"any agent, any cloud: standardized tracing with foundry+opentelemetry","description":"Teams are shipping agents across clouds and frameworks—but telemetry is fragmented. In this demo, see how Microsoft Foundry and OpenTelemetry standards for GenAI tracing bring consistent observability that is agent framework agnostic and cloud agnostic. We’ll walk through the simple setup steps to instrument model and tool calls, quickly diagnose failures, latency, and cost spikes, and close the loop with trace-based evaluation, visualization and optimization.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Problem Overview:</strong> The session opens with greetings and the speaker introducing themselves as a software engineering manager on the Foundry of the Ability team at Microsoft (<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>). They welcome attendees to Day 2 of Build and outline the focus: achieving consistent, high-quality, and actionable agent observability across all frameworks and cloud environments via Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>). Through interactive questions, the speaker identifies challenges with heterogeneous agent setups—different frameworks, hosting stacks, and metrics—across enterprise environments. They introduce Foundry Observability as a way to unify monitoring using OpenTelemetry instrumentation, requiring minimal code changes (<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>).</p>\n\n<p><strong>Foundry Prompt Agent Demonstration:</strong> The next section dives into a hands-on demo using Microsoft Foundry’s portal and playground view. The presenter showcases building a travel expert agent using Foundry’s no-code Prompt Agent (<a class=\"medius-aisummary\" data-starttime=\"00:04:47\">00:04:47</a>). They describe uploading a PDF of travel notes from Xi’an for the agent to reference in generating itineraries. The demo highlights Foundry’s trace view capabilities, showing details like token usage, latency, span counts, and execution trees (<a class=\"medius-aisummary\" data-starttime=\"00:05:56\">00:05:56</a>). With Query metadata visualization and replay controls, the system provides deep insight into RAG patterns and debugging hallucination or retrieval issues (<a class=\"medius-aisummary\" data-starttime=\"00:07:12\">00:07:12</a>). This establishes how Foundry’s observability reveals agent reasoning and behavior step by step.</p>\n\n<p><strong>Multi-Agent and Multi-Cloud Integration:</strong> The speaker invites their colleague Kumar to demonstrate interoperability of agents built outside Foundry. Kumar showcases a Bangalore travel agent built on Google’s ADK running on GCP and another Seattle agent deployed with LangGraph on AWS (<a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>). A Foundry-based orchestrator agent routes user queries intelligently across these specialized agents using OpenTelemetry traces. They demonstrate combined trace visualization showing multiple frameworks and clouds stitched into a unified trace (<a class=\"medius-aisummary\" data-starttime=\"00:11:10\">00:11:10</a>). Queries covering Seattle, Bangalore, and Lisbon show how fallbacks like Copilot SDK handle missing specializations, proving Foundry’s seamless cross-cloud observability (<a class=\"medius-aisummary\" data-starttime=\"00:13:02\">00:13:02</a>).</p>\n\n<p><strong>Core Components and Technical Foundations:</strong> The talk transitions to explaining Foundry’s architecture and OpenTelemetry integration. Foundry supports both no-code Prompt Agents and pro-code Hosted Agents, offering enterprise features such as VM isolation, key identity controls, and long-running operation routines (<a class=\"medius-aisummary\" data-starttime=\"00:13:36\">00:13:36</a>). The second key ingredient is the OpenTelemetry GenAI semantic conventions, ensuring standardized telemetry attributes across agent frameworks. Microsoft contributes to these standards to guarantee compatibility on Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:15:08\">00:15:08</a>). Kumar shows minimal code changes needed to enable auto-instrumentation through Microsoft’s OpenTelemetry Distro and Azure Monitor integration (<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>), enabling agents to instantly emit unified traces regardless of deployment stack.</p>\n\n<p><strong>Operational Monitoring and Evaluation Features:</strong> Foundry’s Monitor and Operate tabs provide real-time operational visibility. Metrics such as latency distributions, token consumption, and error rates help detect anomalies (<a class=\"medius-aisummary\" data-starttime=\"00:18:19\">00:18:19</a>). Automatic alerts trigger for issues like jailbreak attempts or malicious URLs, enhancing security (<a class=\"medius-aisummary\" data-starttime=\"00:19:25\">00:19:25</a>). Evaluation workflows connect directly to traces to identify performance problems — for instance, failed intent adherence when an agent’s output misses a concise comparison task (<a class=\"medius-aisummary\" data-starttime=\"00:21:07\">00:21:07</a>). This illustrates the feedback loop where developers instrument, debug, evaluate, and optimize agents on a single observability plane.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> In closing, the presenters highlight additional Foundry features such as rubric-based evaluators, optimization tools, and support for both OpenTelemetry and OpenInference formats (<a class=\"medius-aisummary\" data-starttime=\"00:22:04\">00:22:04</a>). They encourage checking Foundry’s documentation and demo repository via QR code (<a class=\"medius-aisummary\" data-starttime=\"00:23:20\">00:23:20</a>). The summarizing message unites the session: build any agent, deploy on any cloud, and observe it all through one Foundry observability platform — ensuring consistent, high-quality, actionable monitoring for modern enterprise agent systems (<a class=\"medius-aisummary\" data-starttime=\"00:23:51\">00:23:51</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:50 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/f1e85c0c-4927-4179-bf56-f6ec097af7dc","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/f1e85c0c-4927-4179-bf56-f6ec097af7dc","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/f1e85c0c-4927-4179-bf56-f6ec097af7dc","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/f1e85c0c-4927-4179-bf56-f6ec097af7dc","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/f1e85c0c-4927-4179-bf56-f6ec097af7dc","speakerIds":["1768245294609build26-1776181207688001qvOn","1768245294609build26-1774284998317001piuN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.856Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/f1e85c0c-4927-4179-bf56-f6ec097af7dc","ogTitle":"Any agent, any cloud: Standardized tracing with Foundry+OpenTelemetry","ogDescription":"Teams are shipping agents across clouds and frameworks—but telemetry is fragmented. In this demo, see how Microsoft Foundry and OpenTelemetry standards for GenAI tracing bring consistent observability that is agent framework agnostic and cloud agnostic. We’ll walk through the simple setup steps to instrument model and tool calls, quickly diagnose failures, latency, and cost spikes, and close the loop with trace-based evaluation, visualization and optimization.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998317001piuN","RegistrantKey":"1768245294609build26-1774284998317001piuN"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181207688001qvOn","RegistrantKey":"1768245294609build26-1776181207688001qvOn"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Tracing","logicalValue":"Tracing"},{"displayValue":"Open Telemetry","logicalValue":"Open Telemetry"}]},{"@search.score":1.0,"documentId":"ac6813ba-023d-4c8e-bd49-fd2eecd03fee","startDateTime":"2026-06-03T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jared Rhodes","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290163085001xflH","langLocale":"en-US","sessionId":"1776290163085001xflH","sessionInstanceId":"1776290163085001xflH","sessionCode":"DEM365","sessionCodeNormalized":"EN-US-DEM365","title":"Beyond REST: .NET Apps That Talk to Hardware","sortTitle":"beyond rest: .net apps that talk to hardware","description":"Most app demos stop at REST APIs. This live demo talks to real hardware. I’ll show a .NET app that discovers a device over BLE, exchanges data over NFC or serial, and turns raw device signals into a usable workflow. You’ll leave with a practical pattern for structuring device services, handling platform differences, and deciding when local connectivity beats the cloud.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The video opens with Jared Rhodes greeting the audience and introducing the session titled \"Beyond REST,\" clarifying humorously that the talk is not about web APIs but rather about integrating .NET development with physical hardware <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:18\">00:00:18</a>. He explains that the presentation will explore hardware protocols, testing harnesses, and demos showing how .NET applications can connect hardware components in an agentic workflow <a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>. Jared also mentions feedback from Microsoft on refining the content and introduces Jason, his colleague whose objective is to learn and rationalize hardware development opportunities to his management <a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">01:06</a>.</p>\n\n<p><strong>Exploring .NET, MAUI, and Uno Platform for Hardware Interaction:</strong> Jared transitions to discuss his collaboration with developers like James Montemagno and Sam on .NET MAUI and Uno Platform <a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>. He emphasizes that .NET MAUI applications can run across Windows, macOS, iOS, and Android, making it ideal for cross-platform hardware integrations <a class=\"medius-aisummary\" data-starttime=\"00:01:56\">00:01:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:12\">00:02:12</a>. These platforms can leverage sensors such as accelerometers and GPS, which enrich user experiences beyond web capabilities <a class=\"medius-aisummary\" data-starttime=\"00:02:22\">00:02:22</a>. Jared highlights Uno’s agent-first development tools, which include skills and an MCP server that streamline hardware-based app creation <a class=\"medius-aisummary\" data-starttime=\"00:02:40\">00:02:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>.</p>\n\n<p><strong>Hardware and Real-World Integration Examples:</strong> The presenter links IoT use cases to practical workflows like secure Bluetooth-based authentication that occurs behind QR code scans <a class=\"medius-aisummary\" data-starttime=\"00:03:27\">00:03:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:56\">00:03:56</a>. He describes how NFC simplifies joining Wi-Fi networks and exchanging credentials via physical proximity <a class=\"medius-aisummary\" data-starttime=\"00:04:24\">00:04:24</a>. Hardware-based interactions allow contextual awareness, trust, and identity verification through sensors detecting location and motion, elevating user experiences beyond typical web paradigms <a class=\"medius-aisummary\" data-starttime=\"00:04:38\">00:04:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>. Jared connects this back to the keynote focus on edge computing, explaining that intelligence is moving from the cloud toward on-device inference for efficiency and localized AI responsiveness <a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:35\">00:05:35</a>.</p>\n\n<p><strong>Developing and Demonstrating a Hardware Test Harness:</strong> Jared describes his custom-built test harness, designed to make agentic coding feasible for hardware verification <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:17\">00:07:17</a>. Unlike unit tests or mocks, this harness physically connects to devices for real-world testing, prints line-by-line feedback, and allows deployment automation <a class=\"medius-aisummary\" data-starttime=\"00:07:45\">00:07:45</a>. He shows how it runs via USB and ADB with Android and iOS devices, controlled through Raspberry Pis for distributed testing <a class=\"medius-aisummary\" data-starttime=\"00:08:28\">00:08:28</a>. Jared then launches a live demo, instructing Copilot to execute an NFC test harness involving two phones connected to his basement system <a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:10\">00:12:10</a>. The demonstration shows real-time interaction between the reader and writer devices, reflecting successful AI-driven orchestration of hardware testing, though he notes occasional packet transmission failures typical of real debugging scenarios.</p>\n\n<p><strong>Overview of Key Hardware Protocols – Bluetooth, NFC, and USB:</strong> Following the demo, Jared explains three major communication protocols developers can use: Bluetooth Low Energy (BLE), Near Field Communication (NFC), and USB <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:56\">00:13:56</a>. BLE operates on a service-characteristic model similar to REST APIs, where predefined UUID structures define endpoints for devices such as heart-rate monitors and battery sensors <a class=\"medius-aisummary\" data-starttime=\"00:13:43\">00:13:43</a>. He breaks down the architecture of services, characteristics, and descriptors—all analogs to API paths, methods, and headers <a class=\"medius-aisummary\" data-starttime=\"00:14:17\">00:14:17</a>. The NFC section covers reading and writing tag types, peer-to-peer communication, and Apple’s delayed exposure of the technology to third-party developers <a class=\"medius-aisummary\" data-starttime=\"00:18:17\">00:18:17</a>. Lastly, he introduces the USB binding process—reset, descriptor exchange, configuration, and addressing—and contrasts Android’s openness with Apple's MFi certification requirement for hardware compatibility <a class=\"medius-aisummary\" data-starttime=\"00:21:32\">00:21:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:23\">00:23:23</a>.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> As the presentation concludes <a class=\"medius-aisummary\" data-starttime=\"00:23:35\">00:23:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:57\">00:24:57</a>, Jared emphasizes the creative satisfaction of building applications that interact with the physical world. He encourages developers to experiment with hardware integration through .NET agents and test harnesses, underscoring that physical responses—doors unlocking or sensors activating—create a sense of \"magic\" for users and engineers alike. Viewers are invited to explore his repository for examples of the test harness, continuous integration setups for device-based validation, and references to advanced testing extensions such as FPGA-based simulations in healthcare devices. He closes by thanking attendees and welcoming continued discussion about future agentic workflows and hardware-driven app development. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:30 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0e777bff-486f-41d5-80aa-8a724f6f39f6","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0e777bff-486f-41d5-80aa-8a724f6f39f6","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0e777bff-486f-41d5-80aa-8a724f6f39f6","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0e777bff-486f-41d5-80aa-8a724f6f39f6","speakerIds":["1768245294609build26-1773856399900001oCPe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.904Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0e777bff-486f-41d5-80aa-8a724f6f39f6","ogTitle":"Beyond REST: .NET Apps That Talk to Hardware","ogDescription":"Most app demos stop at REST APIs. This live demo talks to real hardware. I’ll show a .NET app that discovers a device over BLE, exchanges data over NFC or serial, and turns raw device signals into a usable workflow. You’ll leave with a practical pattern for structuring device services, handling platform differences, and deciding when local connectivity beats the cloud.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773856399900001oCPe","RegistrantKey":"1768245294609build26-1773856399900001oCPe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"3e047a7f-b15f-4fce-9066-ede480eaf78f","startDateTime":"2026-06-02T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson, Sunitha Muthukrishna","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018706286001rYPi","langLocale":"en-US","sessionId":"1774018706286001rYPi","sessionInstanceId":"1774018706286001rYPi","sessionCode":"DEM313","sessionCodeNormalized":"EN-US-DEM313","title":"Build agentic apps in minutes with Rayfin and Microsoft Fabric","sortTitle":"build agentic apps in minutes with rayfin and microsoft fabric","description":"Build an enterprise app in minutes with Rayfin, an agent-friendly, code-first backend-as-a-service. In this demo, we show how an AI agent generates a full stack web app with a managed database, authentication, and backend services. Learn how to connect the app to existing data in your Microsoft Fabric data estate, define services in code, and use Fabric’s built-in analytics, BI, and AI capabilities to power insights and intelligent experiences.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"19:30 - 19:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c757302a-b9f5-461f-b57b-b53fdda2d6cd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c757302a-b9f5-461f-b57b-b53fdda2d6cd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c757302a-b9f5-461f-b57b-b53fdda2d6cd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c757302a-b9f5-461f-b57b-b53fdda2d6cd","speakerIds":["1768245294609build26-1776706498871001Ov8x","1768245294609build26-1776443914478001L3uX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.462Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c757302a-b9f5-461f-b57b-b53fdda2d6cd","ogTitle":"Build agentic apps in minutes with Rayfin and Microsoft Fabric","ogDescription":"Build an enterprise app in minutes with Rayfin, an agent-friendly, code-first backend-as-a-service. In this demo, we show how an AI agent generates a full stack web app with a managed database, authentication, and backend services. Learn how to connect the app to existing data in your Microsoft Fabric data estate, define services in code, and use Fabric’s built-in analytics, BI, and AI capabilities to power insights and intelligent experiences.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776443914478001L3uX","RegistrantKey":"1768245294609build26-1776443914478001L3uX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776706498871001Ov8x","RegistrantKey":"1768245294609build26-1776706498871001Ov8x"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM313","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"d918213e-f365-45e2-b1f5-89f58bca7d27","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Lily Du, Umang Sehgal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018702380001r02I","langLocale":"en-US","sessionId":"1774018702380001r02I","sessionInstanceId":"1774018702380001r02I","sessionCode":"DEM334","sessionCodeNormalized":"EN-US-DEM334","title":"Build agents where work happens: chats, channels, and meetings in Microsoft Teams","sortTitle":"build agents where work happens: chats, channels, and meetings in microsoft teams","description":"Agents are moving beyond simple chat experiences. With the Teams SDK, developers can build agents that participate directly in the flow of work across chats, channels, and meetings. In this session, you’ll learn how to leverage Teams capabilities and create agents that automate tasks, surface insights, and take action in context without pulling users out of Teams. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Challenges of Agent Communication:</strong> The video begins by explaining the difficulty of maintaining effective communication between agents and users in a group chat environment compared to one-on-one interactions (<a class=\"medius-aisummary\" data-starttime=\"00:06:17\">00:06:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>). Long-form text that works well in direct chats becomes overwhelming and ineffective in group settings where multiple conversations occur simultaneously. To address this issue, the speaker introduces three key pillars necessary for creating successful collaborative agents and teams: manners, privacy, and polish (<a class=\"medius-aisummary\" data-starttime=\"00:06:38\">00:06:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:46\">00:06:46</a>).</p>\n\n<p><strong>Pillar One - Manners:</strong> The first pillar focuses on teaching agents conversational etiquette similar to human social norms—understanding when to respond, when to create threads, or when to react with emojis (<a class=\"medius-aisummary\" data-starttime=\"00:06:47\">00:06:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:22\">00:07:22</a>). Lily announces several enhancements under \"manners,\" such as emoji reactions, quoted replies, and threaded replies (<a class=\"medius-aisummary\" data-starttime=\"00:08:26\">00:08:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:41\">00:09:41</a>). Emoji reactions allow agents to acknowledge messages nonverbally, quoted replies let them reference older messages unambiguously, and threaded replies keep discussions organized. These features were demonstrated live with the NSYS agent showing polite, contextual responses using emojis and threads in a real group chat scenario (<a class=\"medius-aisummary\" data-starttime=\"00:09:43\">00:09:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>).</p>\n\n<p><strong>Pillar Two - Privacy:</strong> The second pillar highlights privacy and trust as foundations of effective teamwork between humans and agents (<a class=\"medius-aisummary\" data-starttime=\"00:11:11\">00:11:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:17\">00:11:17</a>). New functionality includes targeted messages—enabling agents to privately message individuals without cluttering group chats—and bidirectional private communication, wherein users can send confidential queries to agents using slash commands (<a class=\"medius-aisummary\" data-starttime=\"00:11:46\">00:11:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:10\">00:12:10</a>). This capability is demonstrated in the NSYS system, where the agent privately alerts Umang about a potential issue related to a feature rollout, allowing focused, secure collaboration before posting public updates (<a class=\"medius-aisummary\" data-starttime=\"00:12:16\">00:12:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:34\">00:13:34</a>). The speaker emphasizes that this bidirectional privacy fosters genuine trust and discretion in multi-user environments.</p>\n\n<p><strong>Pillar Three - Polish:</strong> The third pillar covers the presentation and clarity of agent communication to ensure professionalism and readability (<a class=\"medius-aisummary\" data-starttime=\"00:13:46\">00:13:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:04\">00:14:04</a>). Key features include adaptive cards and Markdown support. Adaptive cards transform agent responses into interactive, trackable messages with embedded buttons and structured actions, eliminating long unformatted text blocks (<a class=\"medius-aisummary\" data-starttime=\"00:13:57\">00:13:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:28\">00:14:28</a>). Markdown improves text formatting for better scanning and comprehension, allowing tables, bolding, and code blocks directly in chat (<a class=\"medius-aisummary\" data-starttime=\"00:14:42\">00:14:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:52\">00:14:52</a>). In a follow-up demo, the NSYS agent uses adaptive cards for approval workflows and structured rollback operations, maintaining clarity and accountability while resolving a deployed feature incident (<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:26\">00:17:26</a>).</p>\n\n<p><strong>Demo Recap and Closing Summary:</strong> The final section summarizes all demonstrated capabilities. The speakers recap emoji reactions, quoted replies, and threaded replies for conversation management (<a class=\"medius-aisummary\" data-starttime=\"00:17:31\">00:17:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:48\">00:17:48</a>). They reinforce privacy features like targeted messages and slash commands, followed by adaptive cards and Markdown support for polished communication (<a class=\"medius-aisummary\" data-starttime=\"00:17:50\">00:17:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:01\">00:19:01</a>). Additional features mentioned include source citations, response streaming, feedback buttons, and AI sensitivity labels. The presenters encourage developers to explore the full SDK and new Teams capabilities to build collaborative agents optimized for chats, channels, and meetings (<a class=\"medius-aisummary\" data-starttime=\"00:19:13\">00:19:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>). They conclude with an invitation for further learning through related sessions and documentation, expressing excitement about what builders will create next.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:50 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/b230adf8-4c48-4bb1-9114-6c6bfd355318","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b230adf8-4c48-4bb1-9114-6c6bfd355318","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b230adf8-4c48-4bb1-9114-6c6bfd355318","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b230adf8-4c48-4bb1-9114-6c6bfd355318","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b230adf8-4c48-4bb1-9114-6c6bfd355318","speakerIds":["1768245294609build26-1777487605958001URqq","1768245294609build26-1775594818790001AuRe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.457Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b230adf8-4c48-4bb1-9114-6c6bfd355318","ogTitle":"Build agents where work happens: chats, channels, and meetings in Microsoft Teams","ogDescription":"Agents are moving beyond simple chat experiences. With the Teams SDK, developers can build agents that participate directly in the flow of work across chats, channels, and meetings. In this session, you’ll learn how to leverage Teams capabilities and create agents that automate tasks, surface insights, and take action in context without pulling users out of Teams. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775594818790001AuRe","RegistrantKey":"1768245294609build26-1775594818790001AuRe"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777487605958001URqq","RegistrantKey":"1768245294609build26-1777487605958001URqq"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM334","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"ca1d02e2-f55d-4380-86c0-bbefab267b6a","startDateTime":"2026-06-03T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rajya Laxmi Yellajosyula, Partha Srinivasan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774543682160001e0Zz","langLocale":"en-US","sessionId":"1774543682160001e0Zz","sessionInstanceId":"1774543682160001e0Zz","sessionCode":"DEMSP382","sessionCodeNormalized":"EN-US-DEMSP382","title":"Build AI Apps with Oracle AI Database@Azure, MCP, and GitHub Copilot","sortTitle":"build ai apps with oracle ai database@azure, mcp, and github copilot","description":"Learn to build a secure enterprise AI workflow by combining Oracle Database@Azure, Oracle AI Database, Microsoft Fabric, GitHub Copilot, and Oracle MCP. Explore using MCP and Copilot to orchestrate workflows across Oracle and Fabric using natural language without heavy custom integration. From database provisioning to fraud detection, this demo demonstrates a practical multi-cloud approach for building intelligent, scalable, and enterprise-ready applications.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with an introduction thanking attendees and outlining the topic of discussion — building applications using Oracle Database integrated with Microsoft Azure and MCP servers from both Oracle and Microsoft, utilizing the power of GitHub Copilot <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>. The speaker discusses different developer types, such as traditional app developers, ML engineers, data engineers, and forward deployment engineers, and highlights common challenges faced in accessing and managing data efficiently <a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:00\">01:00:00</a>. Oracle Database on Azure is introduced as a native service available through the Azure Marketplace, enabling seamless integration in development environments <a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:24\">01:01:24</a>.</p>\n\n<p><strong>Challenges and Solutions:</strong> The speaker explains the hurdles developers encounter, particularly with database connectivity, manual wiring, API learning, and schema discovery, along with issues arising from data silos between Oracle and Microsoft systems <a class=\"medius-aisummary\" data-starttime=\"00:01:39\">00:01:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>. To overcome these, Oracle and Microsoft have invested in technologies that streamline integration. Three technologies are highlighted, including Oracle MCP servers accessible via the OCI portal and Microsoft AI Catalog, as well as Fabric MCP servers that enhance connectivity to Microsoft Fabric <a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:17\">00:03:17</a>. GitHub Copilot is positioned as the central tool that unifies these services, simplifying workflow and automation for developers.</p>\n\n<p><strong>Demo 1 – Fraud Detection System:</strong> The first demo demonstrates a fraud detection application <a class=\"medius-aisummary\" data-starttime=\"00:03:22\">00:03:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:08\">00:04:08</a>. Using synthetic customer and account data from an autonomous Oracle AI Database on Azure, the process involves provisioning a database and connecting it through Fabric and Visual Studio. The speaker shares how GitHub Copilot facilitated debugging and setup within minutes, showcasing automation and AI-assisted problem resolution <a class=\"medius-aisummary\" data-starttime=\"00:05:03\">00:05:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:30\">00:05:30</a>. The database tables are created and validated, synthetic data generated automatically, and schema confirmed using Oracle’s SQL CL utility <a class=\"medius-aisummary\" data-starttime=\"00:06:08\">00:06:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:45\">00:07:45</a>. This section emphasizes how natural language prompts through MCP servers quickly translate to SQL queries and how Copilot automates tasks like index tuning and performance optimization that normally require manual intervention <a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>.</p>\n\n<p><strong>Demo 2 – Data Replication and Analytics:</strong> In the second part, data is exported from Oracle to Microsoft Fabric for advanced analytics <a class=\"medius-aisummary\" data-starttime=\"00:09:19\">00:09:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:04\">00:10:04</a>. The process uses secure Azure mechanisms and can leverage Oracle Golden Gate Connector for enterprise-grade replication. Once transferred, the data is verified in Fabric using Fabric MCP server tools and transformed into notebooks ready for ML and Spark analytics workloads <a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>. At this stage, the demo is handed to Raji, who introduces how Oracle AI Database at Azure offers options such as Exadata and Exascale for enterprise operations <a class=\"medius-aisummary\" data-starttime=\"00:11:16\">00:11:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>. Through OCI Golden Gate or Fabric mirroring, real-time Oracle data integration within Fabric is demonstrated using the fictional company Java Media <a class=\"medius-aisummary\" data-starttime=\"00:12:02\">00:12:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>.</p>\n\n<p><strong>Data Engineering and Machine Learning Workflow:</strong> Raji then builds a data pipeline that converts raw (\"bronze\") data to refined (\"gold\") using Fabric MCP server extensions integrated with VS Code <a class=\"medius-aisummary\" data-starttime=\"00:13:31\">00:13:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>. The automation creates notebooks to validate and write data back to the lake house, all executed securely through API connections. The demonstration progresses to machine learning, where an ML engineer uses the Prophet model to forecast Java Media’s viewership metrics <a class=\"medius-aisummary\" data-starttime=\"00:19:03\">00:19:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>. Results are stored in Fabric and displayed as delta tables, emphasizing how MCP servers cut weeks of manual ML model development down to hours using intuitive AI prompts <a class=\"medius-aisummary\" data-starttime=\"00:20:12\">00:20:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>.</p>\n\n<p><strong>Business Insights and Conclusion:</strong> The final section focuses on enabling business-oriented outcomes. Using Fabric MCP servers and REST APIs, developers can create semantic models and data agents to assist business users in interacting with data via conversational insights and Power BI dashboards <a class=\"medius-aisummary\" data-starttime=\"00:21:05\">00:21:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:09\">00:22:09</a>. These tools not only generate forecasts and metrics but also support role-based access controls and integration into Microsoft Stores for broader application use <a class=\"medius-aisummary\" data-starttime=\"00:22:59\">00:22:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:19\">00:23:19</a>. Closing remarks emphasize the unification of Oracle and Microsoft’s ecosystems through the Fabric IQ layer, enabling actionable insights from transactional Oracle data across business workflows <a class=\"medius-aisummary\" data-starttime=\"00:23:23\">00:23:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:10\">00:24:10</a>. Finally, Partha invites attendees to explore more demos in the exhibition booth and mentions available blog resources and giveaways for participants <a class=\"medius-aisummary\" data-starttime=\"00:24:50\">00:24:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:45\">00:25:45</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:30 - 21:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c0c8d8d2-d75e-491a-857e-be570cdbd42f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c0c8d8d2-d75e-491a-857e-be570cdbd42f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c0c8d8d2-d75e-491a-857e-be570cdbd42f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c0c8d8d2-d75e-491a-857e-be570cdbd42f","speakerIds":["1768245294609build26-1768001478850001hB8o","1768245294609build26-1776389175315001yxb8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.841Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Oracle"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c0c8d8d2-d75e-491a-857e-be570cdbd42f","ogTitle":"Build AI Apps with Oracle AI Database@Azure, MCP, and GitHub Copilot","ogDescription":"Learn to build a secure enterprise AI workflow by combining Oracle Database@Azure, Oracle AI Database, Microsoft Fabric, GitHub Copilot, and Oracle MCP. Explore using MCP and Copilot to orchestrate workflows across Oracle and Fabric using natural language without heavy custom integration. From database provisioning to fraud detection, this demo demonstrates a practical multi-cloud approach for building intelligent, scalable, and enterprise-ready applications.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776389175315001yxb8","RegistrantKey":"1768245294609build26-1776389175315001yxb8"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1768001478850001hB8o","RegistrantKey":"1768245294609build26-1768001478850001hB8o"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Demo_Session_Combined_v2_1780202662696001shUd_1780515390734001u5X7.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Vector Embeddings","logicalValue":"Vector Embeddings"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"8519115f-6640-4580-8af9-2b2005c0d5f7","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Brady Gaster, Rong Lu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710086001rTQt","langLocale":"en-US","sessionId":"1774018710086001rTQt","sessionInstanceId":"1774018710086001rTQt","sessionCode":"DEM302","sessionCodeNormalized":"EN-US-DEM302","title":"Build and deploy an Azure app with your agent team","sortTitle":"build and deploy an azure app with your agent team","description":"In 25 minutes, we’ll go from a blank repo to a live Azure AI app—built and deployed by a team of GitHub Copilot agents. Watch Copilot plan architecture, generate code, provision Azure, and deploy live using Copilot Agents and Skills. Leave with a blueprint for your agent driven coding workflow for Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Setup:</strong> The session begins with brief opening remarks and gratitude for the attendees' patience as the presenters prepare for their final demo of the day <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>. The topic of this demo is building and deploying an Azure App Agent. Brady Gaster, recently working with GitHub, and Ron from Microsoft introduce themselves and outline their goal: to showcase how to construct and deploy a customer insights agent system using Azure and GitHub technologies within a focused, 25‑minute demo window <a class=\"medius-aisummary\" data-starttime=\"00:01:09\">00:01:09</a>.</p>\n\n<p><strong>Overview of the Customer Insights Agent:</strong> Ron explains that their demo centers on Zava, an online retailer aiming to understand and act on customer feedback <a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>. The agent system analyzes sentiment—positive or negative—and can automatically trigger responses. For instance, when feedback mentions an app crash or damaged delivery, the agent links to customer data in Microsoft Fabric, identifies orders associated with the feedback, and initiates a replacement process <a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>. The system is designed to enable multiple agents to collaborate to interpret data and take relevant actions autonomously.</p>\n\n<p><strong>Building the Multi-Agent Workflow:</strong> Inside Visual Studio Code, Ron walks through how they developed the system with GitHub Copilot and the Microsoft Agent Framework <a class=\"medius-aisummary\" data-starttime=\"00:04:42\">00:04:42</a>. The first iteration handled simple sentiment analysis, but they expanded into a multi-agent setup consisting of classification, enrichment, triage, and aggregation agents <a class=\"medius-aisummary\" data-starttime=\"00:06:08\">00:06:08</a>. Copilot helped scaffold the entire code base, orchestrating agents in a single workflow definition. Data enrichment connects the system to Microsoft Fabric—allowing one agent to fetch customer and order information through a Foundry toolbox, which serves as an integrated package of Azure AI Search and Fabric connectors accessed via a unified endpoint <a class=\"medius-aisummary\" data-starttime=\"00:08:00\">00:08:00</a>.</p>\n\n<p><strong>Running, Debugging, and Deploying with Foundry:</strong> After setting up locally, Ron demonstrates how the agent runs through the Foundry VS Code extension’s Agent Inspector, a visualization and debugging tool <a class=\"medius-aisummary\" data-starttime=\"00:10:25\">00:10:25</a>. The visual interface displays workflows in real time and enables inspection of data flowing through each agent. The results include drafts of customer communication and sentiment summaries. To scale testing, Copilot automatically adds integrated evaluations based on prior code context <a class=\"medius-aisummary\" data-starttime=\"00:13:09\">00:13:09</a>. Once validated, the agent is deployed as a fully hosted Azure Foundry service for monitoring and log streaming directly inside VS Code, letting Copilot review server logs and resolve errors interactively <a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>.</p>\n\n<p><strong>Front-End Deployment with Squad Framework:</strong> Brady takes over to showcase how he extends Ron’s backend agent by integrating a front end using his multi-agent GitHub Copilot tool called “Squad” <a class=\"medius-aisummary\" data-starttime=\"00:16:10\">00:16:10</a>. Squad creates multiple specialized agents that collaborate efficiently on specific roles instead of competing for tasks <a class=\"medius-aisummary\" data-starttime=\"00:18:38\">00:18:38</a>. Through CLI operations and predefined Azure MCP tools, he sets up an entire Azure Foundry instance and container environment via simple prompts. Brady explains how Squad automatically corrected deployment endpoints, rewrote infrastructure code using Azure skills and the Foundry SDK, and provisioned both backend and frontend containers <a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>. The resulting live application allows users to send feedback from a React interface directly into the Foundry model, securely reusing cloud resources across compute layers <a class=\"medius-aisummary\" data-starttime=\"00:23:24\">00:23:24</a>.</p>\n\n<p><strong>Conclusion and Call to Action:</strong> Wrapping up, Brady shows a Squad-generated report that logs everything the system wrote, deployed, and tested, including the tools and skills used <a class=\"medius-aisummary\" data-starttime=\"00:25:25\">00:25:25</a>. Both presenters highlight the synergy between GitHub Copilot, Azure Foundry, and Squad for building multi-agent systems that operate and document processes autonomously. Attendees are encouraged to explore the Foundry Toolkit for VS Code, Azure MCP tools, GitHub Copilot, and the open-source Squad framework to experiment with agent-based workflows <a class=\"medius-aisummary\" data-starttime=\"00:25:50\">00:25:50</a>. The session closes with thanks and an invitation for viewers to attend the next day’s demonstrations at 9:00 AM <a class=\"medius-aisummary\" data-starttime=\"00:26:03\">00:26:03</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:55:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/29007081-8323-476e-8a69-ee6a106d1cc8","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/29007081-8323-476e-8a69-ee6a106d1cc8","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/29007081-8323-476e-8a69-ee6a106d1cc8","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/29007081-8323-476e-8a69-ee6a106d1cc8","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/29007081-8323-476e-8a69-ee6a106d1cc8","speakerIds":["1768245294609build26-1778720385485001u2mD","1768245294609build26-1775176263055001ZTkl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.789Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/29007081-8323-476e-8a69-ee6a106d1cc8","ogTitle":"Build and deploy an Azure app with your agent team","ogDescription":"In 25 minutes, we’ll go from a blank repo to a live Azure AI app—built and deployed by a team of GitHub Copilot agents. Watch Copilot plan architecture, generate code, provision Azure, and deploy live using Copilot Agents and Skills. Leave with a blueprint for your agent driven coding workflow for Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775176263055001ZTkl","RegistrantKey":"1768245294609build26-1775176263055001ZTkl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778720385485001u2mD","RegistrantKey":"1768245294609build26-1778720385485001u2mD"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"6e540d94-9fa1-49ce-8e48-8776e38dfe51","startDateTime":"2026-06-02T20:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Darren Oberst, Meghana Rao, Morris Novello","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774391893433001yV7X","langLocale":"en-US","sessionId":"1774391893433001yV7X","sessionInstanceId":"1774391893433001yV7X","sessionCode":"DEMSP380","sessionCodeNormalized":"EN-US-DEMSP380","title":"Build automated agents using optimized AI Foundry models on Snapdragon","sortTitle":"build automated agents using optimized ai foundry models on snapdragon","description":"Build performant, secure local agents customized for enterprise use cases on Snapdragon X Series PCs. Using GenAI models optimized for the Qualcomm Hexagon NPU architecture, this session empowers developers to build automated, scheduled agentic AI workflows that unlock true NPU acceleration at the edge.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Motivation:</strong> At the start of the session <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>, the presenters introduce the goal of demonstrating how to build automated AI agents using optimized models available in Microsoft Foundry for Snapdragon PCs. They highlight how many enterprise workflows today rely on cloud-based chat models that respond to prompts, but such setups do not easily handle complex, recurring, or multi-step workflows. The discussion <a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a> emphasizes the importance of running models locally, especially in secure enterprise environments, and introduces the idea of scheduled, prompt-free, automated agents powered by LLMware’s Model HQ and Windows ML APIs for Snapdragon-based devices.</p>\n\n<p><strong>Learning Goals and Enterprise Use Case:</strong> The speakers outline the four key takeaways <a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:11\">00:02:11</a>: creating multi-step workflows powered by small language models, building industry-specific automation, achieving optimized on-device performance via the Snapdragon NPU, and setting scheduled automation runs. To illustrate this, they choose the Jira workflow as a concrete example <a class=\"medius-aisummary\" data-starttime=\"00:02:14\">00:02:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:03\">00:03:03</a>, discussing how enterprises use multiple Jira instances to track feature requests, bugs, and user stories. LLMware’s Model HQ assists in integrating and extracting key insights from these siloed datasets using a drag-and-drop, no-code design interface, simplifying creation of agentic workflows on structured data.</p>\n\n<p><strong>Architecture and Demo Preview:</strong> Before the live demonstration, a brief explanation is given of the Model HQ architecture <a class=\"medius-aisummary\" data-starttime=\"00:03:36\">00:03:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>. At the center sits the no-code application integrated with Microsoft Foundry and external model repositories, running on Snapdragon PCs through Windows ML and Onnx runtime APIs supported by Qualcomm execution providers for optimal local performance. A preview video <a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a> shows what the live demo will cover, giving participants a quick overview of building and running automated workflows. Technical issues with toggling screens are briefly encountered but later resolved as the presenters transition to the live demo portion.</p>\n\n<p><strong>Live Demonstration of Agent Creation:</strong> The main presenter, Darren, begins walking through the live configuration <a class=\"medius-aisummary\" data-starttime=\"00:06:51\">00:06:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>. He integrates three components: the Jira API, Windows Foundry Local, and an email client to automate the end-to-end process. Using the Foundry Local NPU model optimized for Snapdragon X2, Darren shows real-time summarization of Jira issues and highlights that the workflow can be scheduled to deliver daily reports directly to inboxes without token costs or cloud dependencies. The demo illustrates Model HQ's local execution speed and privacy advantages due to on-device processing. The presenter demonstrates drag-and-drop workflow creation <a class=\"medius-aisummary\" data-starttime=\"00:10:24\">00:10:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:05\">00:11:05</a> and emphasizes how even non-technical users can build workflows, leveraging Snapdragon’s NPU for accelerated inference.</p>\n\n<p><strong>Programmatic Extension and Integration:</strong> The speakers continue to showcase how no-code agents can be extended programmatically <a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>, describing the ability to expose functionalities as Windows services accessible via API endpoints. By enabling local host access, users can call agents asynchronously, get execution IDs, retrieve outputs, and schedule operations without any cloud dependency. The SDK demonstration <a class=\"medius-aisummary\" data-starttime=\"00:15:03\">00:15:03</a> shows how to invoke and manage agent processes programmatically, streamlining automation for repeated enterprise tasks. The team demonstrates pulling zip folders of outputs and explains that these agents can later be deployed at scale on servers, transforming local setups into distributed systems for broader organizational use.</p>\n\n<p><strong>Conclusion and Q&A:</strong> In the final section <a class=\"medius-aisummary\" data-starttime=\"00:20:02\">00:20:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:23\">00:24:23</a>, audience questions focus on scalability and hardware optimization. The presenters confirm support for over 200 models, including cloud calls for complex reasoning or multimodal tasks that can be selectively integrated with local workflows, achieving a hybrid setup. They highlight Snapdragon X2’s advantages—up to 80 TOPS acceleration—offering higher-speed local inference on large models. Concluding, they invite attendees to the Qualcomm booth to explore further use cases and emphasize the benefits of embracing local AI for cost efficiency, privacy, and performance. The session closes with encouragement to try hands-on agent building and extend the innovation from hardware through no-code software capabilities in Snapdragon-powered AI PCs.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"20:40 - 21:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:05:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/5598fc68-e815-4c2e-b5fc-b9bd4ad90c22","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/5598fc68-e815-4c2e-b5fc-b9bd4ad90c22","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/5598fc68-e815-4c2e-b5fc-b9bd4ad90c22","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/5598fc68-e815-4c2e-b5fc-b9bd4ad90c22","speakerIds":["1768245294609build26-1776715652642001HevV","1768245294609build26-1776715202121001H93x","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.637Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/5598fc68-e815-4c2e-b5fc-b9bd4ad90c22","ogTitle":"Build automated agents using optimized AI Foundry models on Snapdragon","ogDescription":"Build performant, secure local agents customized for enterprise use cases on Snapdragon X Series PCs. Using GenAI models optimized for the Qualcomm Hexagon NPU architecture, this session empowers developers to build automated, scheduled agentic AI workflows that unlock true NPU acceleration at the edge.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776715202121001H93x","RegistrantKey":"1768245294609build26-1776715202121001H93x"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776715652642001HevV","RegistrantKey":"1768245294609build26-1776715652642001HevV"},{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/DEMSP380%20-%20MSFT%20Build%202026%20-%20Speakerships%20-%20Demo%20Session%20SD3.0_1780006041110001V11e_1780512502912001wdnK.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"}]},{"@search.score":1.0,"documentId":"c1851708-4087-4a63-9210-85c59c37a5cb","startDateTime":"2026-06-02T21:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeff Vestal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775146286110001hUDH","langLocale":"en-US","sessionId":"1775146286110001hUDH","sessionInstanceId":"1775146286110001hUDH","sessionCode":"DEMSP385","sessionCodeNormalized":"EN-US-DEMSP385","title":"Build context-aware agents using GitHub Copilot, Elastic, and Azure AI","sortTitle":"build context-aware agents using github copilot, elastic, and azure ai","description":"Your AI reviewer has amnesia; it knows your code, but forgets the 2 AM outage from last month. We’ll demo giving Copilot a memory. Using Elastic Agent Builder and Workflows, we build an agent that reviews PRs against production telemetry and incident history. See it in action: an agent identifies a regression pattern, cites a past incident via OTel traces in Elasticsearch, and helps a dev fix it in VS Code via Copilot Agent Mode and the Elastic MCP server. Powered by Azure AI Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Objective:</strong> The presenter begins by greeting the audience and gauging familiarity with Elasticsearch <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>. The session focuses on using Elasticsearch within a code review workflow to automatically identify whether a new pull request (PR) might reintroduce problems from previous incidents <a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>. Instead of only validating that code compiles or passes tests, the goal is to determine if a change could recreate past production issues. This enhanced review process ties together postmortem incident data, open telemetry traces, and Elasticsearch intelligence to flag potentially risky changes automatically before merging <a class=\"medius-aisummary\" data-starttime=\"00:00:26\">00:00:26</a>.</p>\n\n<p><strong>System Workflow Overview:</strong> The presentation transitions to explain how the automation works. When developers push a PR, standard GitHub Actions trigger the usual validation steps but also invoke a new Elastic Workflow <a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>. This, in turn, calls Elastic Agent Builder—the platform’s agent execution environment. The agent conducts semantic searches in Elasticsearch, comparing the proposed code changes against telemetry data and postmortems for incident similarities <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>. When matches are found, it comments on the PR with context from these findings, suggesting whether to proceed or adjust. Further integration with tools like VS Code and Copilot allows developers to interact with recommendations and automatically apply safe fixes guided by the system <a class=\"medius-aisummary\" data-starttime=\"00:01:42\">00:01:42</a>.</p>\n\n<p><strong>Demonstration: Pull Request Detection and Feedback:</strong> The presenter demonstrates with a sample e-commerce app, Wayfinder, where a small refactoring in the inventory reservation function leads to a code review run <a class=\"medius-aisummary\" data-starttime=\"00:02:24\">00:02:24</a>. Though syntactically valid, the Elastic PR agent identifies that the proposed change reintroduces a race condition previously responsible for production gateway timeouts on April 7th <a class=\"medius-aisummary\" data-starttime=\"00:04:12\">00:04:12</a>. The feedback comment includes detailed telemetry evidence showing latency spikes, specific root causes, and example traces. The agent provides not only a warning but a recommended fix and coverage of the incident history, empowering the developer to adjust code confidently before merging.</p>\n\n<p><strong>Elastic Agent Builder and Semantic Tools:</strong> The walkthrough continues inside Kibana’s agent interface <a class=\"medius-aisummary\" data-starttime=\"00:06:20\">00:06:20</a>. The speaker explains that agents are configured with “skills” (contextual knowledge like recognizing race conditions) and “tools” (functional access to Elasticsearch search, semantic matching, and ESQL queries) <a class=\"medius-aisummary\" data-starttime=\"00:07:56\">00:07:56</a>. Using a workflow, the PR review can automatically collect environment variables from GitHub and feed them into Elasticsearch for correlation against vectors generated from incident embeddings <a class=\"medius-aisummary\" data-starttime=\"00:09:52\">00:09:52</a>. Unlike keyword matching, the system employs vector-based semantic search using model embeddings, enabling it to reason about contextual code issues and map them against historic failure patterns.</p>\n\n<p><strong>Developer Interaction through Queries and Copilot:</strong> The presenter then demonstrates how developers can directly converse with the agent for pre-PR analysis <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>. By asking whether certain endpoints or concurrency modifications match known incident patterns, the agent retrieves correlated data, root-cause explanations, and postmortem summaries, often pinpointing exact production failures <a class=\"medius-aisummary\" data-starttime=\"00:13:06\">00:13:06</a>. Through integrated MCP servers, VS Code, and GitHub Copilot, developers can keep their workflow inside the IDE. Copilot can automatically apply code updates based on the agent’s recommendations, as shown when it refactors the database statement to prevent threading issues and confirms correctness through local stress testing <a class=\"medius-aisummary\" data-starttime=\"00:16:15\">00:16:15</a>.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> The session concludes by recapping how integrating postmortems, telemetry, and RCA data into Elasticsearch allows engineering teams to move from reactive to proactive code review <a class=\"medius-aisummary\" data-starttime=\"00:17:08\">00:17:08</a>. Developers can inject varying levels of observability data—whether production incidents or dev-stage metrics—to create agents that foresee repeat failures. The outcome is a more resilient, context-aware review process that safeguards production stability automatically within developers’ daily tools. The presenter thanks the audience and invites them to visit the Elastic booth for hands-on demonstrations and giveaways <a class=\"medius-aisummary\" data-starttime=\"00:17:46\">00:17:46</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:10 - 21:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:35:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d517366a-a8ce-4b50-9047-aeb5e8309c2b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d517366a-a8ce-4b50-9047-aeb5e8309c2b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d517366a-a8ce-4b50-9047-aeb5e8309c2b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d517366a-a8ce-4b50-9047-aeb5e8309c2b","speakerIds":["1768245294609build26-1776975280216001ylMK"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.844Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Elastic"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d517366a-a8ce-4b50-9047-aeb5e8309c2b","ogTitle":"Build context-aware agents using GitHub Copilot, Elastic, and Azure AI","ogDescription":"Your AI reviewer has amnesia; it knows your code, but forgets the 2 AM outage from last month. We’ll demo giving Copilot a memory. Using Elastic Agent Builder and Workflows, we build an agent that reviews PRs against production telemetry and incident history. See it in action: an agent identifies a regression pattern, cites a past incident via OTel traces in Elasticsearch, and helps a dev fix it in VS Code via Copilot Agent Mode and the Elastic MCP server. Powered by Azure AI Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776975280216001ylMK","RegistrantKey":"1768245294609build26-1776975280216001ylMK"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/msft-build-2026-submission_1777920789128001JWh4_1780512427007001NX2a.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"}]},{"@search.score":1.0,"documentId":"c33fd77a-c0ea-4cfa-ac42-130ded6d1e5a","startDateTime":"2026-06-03T18:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Henk Boelman","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701742001rOMi","langLocale":"en-US","sessionId":"1774018701742001rOMi","sessionInstanceId":"1774018701742001rOMi","sessionCode":"DEM330","sessionCodeNormalized":"EN-US-DEM330","title":"Build multimodal agents that reason, interact, and take action","sortTitle":"build multimodal agents that reason, interact, and take action","description":"Build a real-time agent with a conversational avatar using the Voice Live API in Microsoft Foundry, no slides, just live coding. See how Voice Live unifies speech-to-text, model reasoning, and text-to-speech, then wire it into an agent that can call tools via MCP to take real actions. Walk away with reusable patterns for voice-first, tool-using agents with expressive avatars.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a> Heng Bullmann, a Principal Developer Advocate at Microsoft, welcomes viewers to a session focused on building multimodal agents that can reason, interact, and take action. He introduces the concept of digital agents as the next evolution in AI applications and mentions how most developer conversations are increasingly centered around them. To demonstrate this, he presents \"Hank’s Travel Agency,\" a voice agent that can communicate naturally, search for flights and hotels, and respond interactively in real time.</p>\n\n<p><strong>Interactive Voice Agent Demo:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:00:29\">00:00:29</a>, the presenter shows a real-time demo of the travel assistant. The voice agent converses with Heng, managing his trip to Microsoft Build in San Francisco. It successfully books a flight departing from Amsterdam on May 28, 2026, confirming pricing and times (<a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:39\">00:01:39</a>) and offering hotel suggestions within his budget (<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:34\">00:02:34</a>). It finalizes both travel and lodging details conversationally, ending with flight returns and confirmation of the complete trip. At <a class=\"medius-aisummary\" data-starttime=\"00:03:26\">00:03:26</a>, Heng recaps how he talked to the assistant as if it were human, while it listened, adapted, and carried out tasks in real time.</p>\n\n<p><strong>Technology Behind the Voice Live API:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:03:44\">00:03:44</a>, Heng explains that the demo runs on Microsoft’s Voice Live API—a unified interface that merges speech-to-text and text-to-speech for seamless, real-time communication. It automatically handles listening, turn-taking, and interruptions. The demo uses the new Neural HD V3 voice “Serena” and Maya Transcribe 1 integration for multilingual speech input. The system’s lifelike avatar “Selena” runs over WebRTC, enabling browser-based real-time audio-visual interaction.</p>\n\n<p><strong>Building an Agent in Microsoft Foundry:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:04:43\">00:04:43</a>, Heng walks through how such an agent is created in Microsoft Foundry. Step one (<a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>) involves setting up a “Prompt Agent,” defining its model and conversational instructions, and connecting it to the Contoso Travel MCP server. The agent is granted access to external services automatically, enabling it to handle bookings and queries. He demonstrates testing the text-only version directly in Foundry, confirming it can already perform intelligent actions before voice integration.</p>\n\n<p><strong>Customizing Voice and Avatar:</strong> Heng then introduces personalization options beginning at <a class=\"medius-aisummary\" data-starttime=\"00:07:06\">00:07:06</a>. He describes the Neural HD voices that dynamically adapt tone and context, the latest MyVoice 2 model for expressive speech, and the use of Personal Voice cloning (<a class=\"medius-aisummary\" data-starttime=\"00:08:12\">00:08:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:59\">00:09:59</a>) created by recording a consent statement and short sample in Foundry. This produces a unique recreated voice. He then adds a matching avatar (<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:15\">00:11:15</a>), describing the creation of lifelike visuals using a single photograph and a consent video to train a talking head model—resulting in realistic, expressive avatars.</p>\n\n<p><strong>Final Integration and Conclusion:</strong> In the closing section (<a class=\"medius-aisummary\" data-starttime=\"00:11:54\">00:11:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:59\">00:15:59</a>), Heng combines everything: his personalized voice and avatar interacting through the same agent system, now booking a new trip to New York. The reconstruction shows how the multimodal agent functions autonomously from flight to hotel selection while reflecting Heng’s own identity visually and vocally. He summarizes that developers can create such agents through Microsoft Foundry and Voice Live by focusing on intent and user experience, since the complex conversational handling is automated. He concludes by encouraging viewers to explore the demo at the provided URL and create multimodal agents that are uniquely their own.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:20 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4ae2998c-c3ca-44ab-b896-41047cd9c688","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4ae2998c-c3ca-44ab-b896-41047cd9c688","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4ae2998c-c3ca-44ab-b896-41047cd9c688","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4ae2998c-c3ca-44ab-b896-41047cd9c688","speakerIds":["1768245294609build26-1754963585693001VLHf"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.457Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4ae2998c-c3ca-44ab-b896-41047cd9c688","ogTitle":"Build multimodal agents that reason, interact, and take action","ogDescription":"Build a real-time agent with a conversational avatar using the Voice Live API in Microsoft Foundry, no slides, just live coding. See how Voice Live unifies speech-to-text, model reasoning, and text-to-speech, then wire it into an agent that can call tools via MCP to take real actions. Walk away with reusable patterns for voice-first, tool-using agents with expressive avatars.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1754963585693001VLHf","RegistrantKey":"1768245294609build26-1754963585693001VLHf"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"}]},{"@search.score":1.0,"documentId":"f58a4af3-c189-40f7-ab0b-2737fb27b0e0","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Srikumar (Sri) Nair","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717046001rNhm","langLocale":"en-US","sessionId":"1774018717046001rNhm","sessionInstanceId":"1774018717046001rNhm","sessionCode":"DEM340","sessionCodeNormalized":"EN-US-DEM340","title":"Build work-ready agents with Foundry + Work IQ, govern with Agent 365","sortTitle":"build work-ready agents with foundry + work iq, govern with agent 365","description":"An agent that knows nothing about your work is only half an agent. In this demo, we connect a Foundry agent to Work IQ via MCP, giving it real context about your work and ecosystem, governed by Agent 365. See it run under agent identity, watch Work IQ tools invoked at runtime, and get full observability on every call. Then we run side-by-side evals context-aware vs. context-blind, with scores that prove the difference. Leave knowing how to ground, govern, and evaluate your Foundry agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with an energetic greeting from Sri, the Group Product Manager for Agent 365, addressing developers at <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>. Sri introduces himself as responsible for helping teams build AI agents while ensuring they are governable within enterprise systems. He explains the key enterprise challenge: IT administrators want control and transparency over what AI is doing within their systems, only approving what meets compliance and governance needs. After playful engagement with the audience about familiarity with Foundry IQ, Work IQ, and Agent 365, he outlines the session plan — a rapid demo covering Work IQ, agent creation, evaluations, security, governance, and a final surprise topic <a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>.</p>\n\n<p><strong>Understanding Context and Work IQ:</strong> Sri stresses that “context is key” when building effective AI agents <a class=\"medius-aisummary\" data-starttime=\"00:01:27\">00:01:27</a>. He details how context — stored across enterprise data sources such as SQL, SAP, emails, and chats — empowers agents to make intelligent, relevant decisions (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>). Work IQ consolidates these contexts from Microsoft 365 assets like email, Teams, and SharePoint into a single, secure intelligence layer that agents leverage for responses. Sri explains that Work IQ includes data, context, skills, and tools, with the Work IQ MCP API unlocking secure access to contextual enterprise data. He announces general availability targeting mid-June <a class=\"medius-aisummary\" data-starttime=\"00:03:11\">00:03:11</a>, promising full security compliance so agents only see appropriate data per identity permissions.</p>\n\n<p><strong>Demonstrating Foundry IQ and Integration with Work IQ:</strong> Transitioning to the demo <a class=\"medius-aisummary\" data-starttime=\"00:05:29\">00:05:29</a>, Sri builds a playful “Agent 007” in the Foundry UI. He demonstrates how simple it is to attach Work IQ to the agent catalog <a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>, automatically connecting with all enterprise context sources. Then, he introduces Foundry IQ as the system’s knowledge backbone <a class=\"medius-aisummary\" data-starttime=\"00:07:39\">00:07:39</a>—a tool that converts enterprise data stored in blob containers into searchable, vector-indexed knowledge through Azure AI Search. He shows how connecting Foundry IQ to an agent gives instant access to files like pricing guides or product specs, meaning the agent can deliver grounded, context-aware answers. By combining Work IQ for enterprise context with Foundry IQ for deep knowledge, Sri demonstrates how easily developers can enrich agents with secure, compliant information <a class=\"medius-aisummary\" data-starttime=\"00:10:09\">00:10:09</a>.</p>\n\n<p><strong>Evaluation, Testing, and Publishing:</strong> Sri introduces “evals,” a built-in Foundry capability to test and validate agents before deployment <a class=\"medius-aisummary\" data-starttime=\"00:12:07\">00:12:07</a>. Evals assess agents for metrics such as task completion, customer satisfaction, coherence, and groundedness — ensuring that answers remain faithful to enterprise knowledge bases and not hallucinated <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. After uploading sample evaluation data, he shows how the system generates quantitative performance scores, identifying readiness for production deployment. Once validated, the agent can be published directly to the company registry or Microsoft Teams channels. Foundry automatically provisions a bot service and agent ID, drastically simplifying distribution processes <a class=\"medius-aisummary\" data-starttime=\"00:16:07\">00:16:07</a>.</p>\n\n<p><strong>Agent 365: Governance, Observability, and Security:</strong> Switching perspectives to that of an IT administrator <a class=\"medius-aisummary\" data-starttime=\"00:16:40\">00:16:40</a>, Sri showcases Agent 365’s governance tools. The admin interface allows viewing active agents, their risk posture, usage analytics, and required approvals. Each agent is automatically registered with a unique “agent identity,” carrying principal permissions similar to a human user for fine-grained access control. Sri explains that Agent 365 bridges a 52% governance gap between agents deployed in production and those compliant with enterprise standards <a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>. The system integrates security features through Intune for sandboxed environments, Defender for prompt injection protection, and policies for data labeling, access restriction, and exfiltration blocking. These capabilities ensure that agents follow enterprise policies rather than arbitrary model behavior.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> The presentation wraps up <a class=\"medius-aisummary\" data-starttime=\"00:25:22\">00:25:22</a> by demonstrating the agent registry map view, which visualizes all AI agents across platforms including third-party systems such as Databricks and AWS. IT admins can monitor local agents like GitHub Copilot, set isolation policies, and enforce compliant workspace execution <a class=\"medius-aisummary\" data-starttime=\"00:23:20\">00:23:20</a>. Tool governance in Agent 365 ensures that MCP connectors, APIs, and other agent plugins adhere to strict regulations. Sri ends by reinforcing the dual mission: build intelligent, context-rich agents in Foundry while maintaining governance and security via Agent 365. The combined ecosystem delivers observability, compliance, and protection across the full lifecycle of enterprise AI agents <a class=\"medius-aisummary\" data-starttime=\"00:25:38\">00:25:38</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:00 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4ce5f64d-3565-4d14-8010-32ad80dc7524","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4ce5f64d-3565-4d14-8010-32ad80dc7524","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4ce5f64d-3565-4d14-8010-32ad80dc7524","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4ce5f64d-3565-4d14-8010-32ad80dc7524","speakerIds":["1768245294609build26-1776362105831001L2PN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.856Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4ce5f64d-3565-4d14-8010-32ad80dc7524","ogTitle":"Build work-ready agents with Foundry + Work IQ, govern with Agent 365","ogDescription":"An agent that knows nothing about your work is only half an agent. In this demo, we connect a Foundry agent to Work IQ via MCP, giving it real context about your work and ecosystem, governed by Agent 365. See it run under agent identity, watch Work IQ tools invoked at runtime, and get full observability on every call. Then we run side-by-side evals context-aware vs. context-blind, with scores that prove the difference. Leave knowing how to ground, govern, and evaluate your Foundry agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776362105831001L2PN","RegistrantKey":"1768245294609build26-1776362105831001L2PN"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Work IQ","logicalValue":"Work IQ"},{"displayValue":"Context Engineer","logicalValue":"Context Engineer"},{"displayValue":"Grounding","logicalValue":"Grounding"}]},{"@search.score":1.0,"documentId":"b3f520bf-85dc-41e3-9de8-41b2c225d3a3","startDateTime":"2026-06-03T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alexander Wachtel","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779813985656001idOZ","langLocale":"en-US","sessionId":"1779813985656001idOZ","sessionInstanceId":"1779813985656001idOZ","sessionCode":"DEM362-R1","sessionCodeNormalized":"EN-US-DEM362-R1","title":"Building a Multi-Agent Workflow in Microsoft Fabric","sortTitle":"building a multi-agent workflow in microsoft fabric","description":"Learn how Microsoft Fabric and the 3 IQs, Work IQ, Fabric IQ, and Foundry IQ, enable dynamic, context‑aware multi‑agent workflows. See how semantic models, lakehouses, and pipelines form agent‑ready foundations, and how specialized agents collaborate across planning, retrieval, summarization, and execution. Discover how natural language, automation, and orchestration power real multi‑agent scenarios in Microsoft Fabric.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with Alex, an MVP and trainer from Germany, welcoming participants and apologizing for his accent before introducing the topic of the talk <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>. He explains that this is the second session of the day and will focus on a demonstration related to Microsoft’s Fabric IQ, Foundry IQ, and Work IQ environments <a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>. These three areas represent interconnected components in data management and AI operations across organizations. Alex emphasizes the goal of showing how data agents interact across systems to deliver insights and automation efficiently before moving into the detailed demo portion.</p>\n\n<p><strong>Understanding the IQ Frameworks:</strong> Alex describes Microsoft’s various IQ systems: Work IQ, Fabric IQ, and Foundry IQ <a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>. Work IQ manages organizational data such as emails, meetings, and documents; Fabric IQ focuses on ontological company structures; and Foundry IQ provides governance and access policies. He explains how these modules interoperate through Microsoft Copilot and data agents. The ontology used in Fabric IQ helps represent company knowledge and data relationships in a more semantic manner than traditional SQL-based tables. Foundry IQ further ensures compliance and controlled data access via permission management. Throughout, Alex stresses that these systems are designed to transform traditional data silos into intelligent, interconnected ecosystems that support AI-driven querying and reasoning.</p>\n\n<p><strong>Demo: Data Agents and the No-Black-Box Principle:</strong> The demonstration begins with Alex switching roles between different personas—himself as the Foundry administrator, Hana as the Fabric data engineer, and Rafael as the compliance officer <a class=\"medius-aisummary\" data-starttime=\"00:03:11\">00:03:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>. He recalls earlier approaches using Azure OpenAI Studio to create agents that accessed company data but notes the absence of transparency (the \"black box\" issue). Microsoft’s new system now eliminates that problem, allowing visibility into agent queries and data usage. Alex demonstrates Work IQ agents querying company emails and documents via Copilot, which internally triggers services like Work IQ Mail MCP servers to process the request <a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a>. He also showcases how user approval and security validation steps are integrated whenever external connections occur, reinforcing data trust and compliance.</p>\n\n<p><strong>Applying Fabric IQ and Ontologies:</strong> Shifting focus to Fabric IQ, Alex highlights its ontology-driven approach to data management, where information is structured as nodes and relationships rather than mere datasets <a class=\"medius-aisummary\" data-starttime=\"00:12:14\">00:12:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:12\">00:14:12</a>. He defines how data flows from bronze through silver to gold layers, aligning with Microsoft’s data maturity model. Using examples like retail store sales, Alex shows how data agents query the “lakehouse” layer to retrieve real-time insights <a class=\"medius-aisummary\" data-starttime=\"00:13:52\">00:13:52</a>. He compares this modern integration with older setups, emphasizing how semantic models and ontologies now simplify relationships for AI agents. Alex then explores publishing these data agents to Copilot or Foundry platforms to extend their use. This modular publishing model allows both technical and non-technical users to access structured intelligence through natural language queries, turning Fabric IQ into a central hub for enterprise knowledge.</p>\n\n<p><strong>Agent SDK, Evaluation, and Governance:</strong> Alex moves deeper into technical territory by introducing the Data Agent SDK <a class=\"medius-aisummary\" data-starttime=\"00:19:45\">00:19:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:01\">00:21:01</a>. He demonstrates how developers can connect Azure OpenAI endpoints with access tokens and evaluate their agents’ responses programmatically. This SDK enables detailed testing and validation, ensuring agents provide accurate and compliant answers. He underscores that while the Fabric portal allows publishing and simple configuration, the SDK offers higher control through code-level evaluation. With this approach, engineers can detect errors and validate correctness before users interact with the system. He also mentions ongoing orchestration limitations within Fabric but notes that pipelines can be used to chain workflows and evaluate agent performance automatically.</p>\n\n<p><strong>Conclusion and Governance View:</strong> In closing, Alex introduces “Agent 365,” a unified view that tracks all operational agents across Foundry, Fabric, Copilot Studio, and Microsoft 365 <a class=\"medius-aisummary\" data-starttime=\"00:23:45\">00:23:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:45\">00:25:45</a>. This governance tool allows administrators to see which agents exist, their publishers, activity history, and interactions, including sensitive data access. The Activity Explorer provides time-based insights into user queries and AI actions, ensuring that compliance officers like Rafael can oversee data usage transparently. Alex emphasizes that the days of “black box AI” are gone—now, every data retrieval, analysis, and exchange is auditable. He concludes by thanking the audience for attending and encourages them to explore post-Ignite templates and repository materials to create their own agents based on the demonstrated framework <a class=\"medius-aisummary\" data-starttime=\"00:25:46\">00:25:46</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:30 - 22:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d48b1a04-b191-438e-963b-2b3b03577cb6","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d48b1a04-b191-438e-963b-2b3b03577cb6","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d48b1a04-b191-438e-963b-2b3b03577cb6","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d48b1a04-b191-438e-963b-2b3b03577cb6","speakerIds":["1768245294609build26-1760032715730001qqSb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.238Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d48b1a04-b191-438e-963b-2b3b03577cb6","ogTitle":"Building a Multi-Agent Workflow in Microsoft Fabric","ogDescription":"Learn how Microsoft Fabric and the 3 IQs, Work IQ, Fabric IQ, and Foundry IQ, enable dynamic, context‑aware multi‑agent workflows. See how semantic models, lakehouses, and pipelines form agent‑ready foundations, and how specialized agents collaborate across planning, retrieval, summarization, and execution. Discover how natural language, automation, and orchestration power real multi‑agent scenarios in Microsoft Fabric.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1760032715730001qqSb","RegistrantKey":"1768245294609build26-1760032715730001qqSb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"bce1ce24-4594-461a-99ac-8c4a8006f15c","startDateTime":"2026-06-03T19:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alexander Wachtel","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776288171535001NKxq","langLocale":"en-US","sessionId":"1776288171535001NKxq","sessionInstanceId":"1776288171535001NKxq","sessionCode":"DEM362","sessionCodeNormalized":"EN-US-DEM362","title":"Building a Multi-Agent Workflow in Microsoft Fabric","sortTitle":"building a multi-agent workflow in microsoft fabric","description":"Learn how Microsoft Fabric and the 3 IQs, Work IQ, Fabric IQ, and Foundry IQ, enable dynamic, context‑aware multi‑agent workflows. See how semantic models, lakehouses, and pipelines form agent‑ready foundations, and how specialized agents collaborate across planning, retrieval, summarization, and execution. Discover how natural language, automation, and orchestration power real multi‑agent scenarios in Microsoft Fabric.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The session begins with a friendly welcome and acknowledgment of limited space before outlining the topics: agents, Microsoft Fabric, and several \"IQ\" components (<a class=\"medius-aisummary\" data-starttime=\"00:00:10\">00:00:10</a>). The speaker briefly mentions connecting on LinkedIn for follow-up questions and sets the scene for exploring how orchestration works within data pipelines using notebooks and integrating agents across Foundry IQ and Worker Queue (<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>). The goal of the presentation is to connect concepts of Fabric IQ, Worker IQ, and Foundry IQ under Microsoft's growing \"agent\" framework used for enterprise AI and data orchestration.</p>\n\n<p><strong>Concepts Behind Microsoft IQs:</strong> The next section establishes context around Microsoft’s IQ infrastructure, describing Worker IQ, Fabric IQ, and Foundry IQ as essential layers providing agents with context and knowledge (<a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>). The speaker recalls earlier chatbot work during a PhD, highlighting limitations of static knowledge bases and failed answers outside predefined scope — a lesson carried forward into today’s dynamic AI architectures. Worker Queue represents everyday Microsoft 365 data like emails, meetings, and files, while Fabric IQ handles organizational business data, such as performance and sales metrics (<a class=\"medius-aisummary\" data-starttime=\"00:02:40\">00:02:40</a>). Foundry IQ, on the other hand, serves as a unified knowledge base combining diverse internal and external sources to support intelligent agents within a trusted environment.</p>\n\n<p><strong>Team Roles and Historical Challenges:</strong> The speaker introduces the three-person collaboration model within Microsoft Build projects: the agent designer, data engineer, and compliance officer (<a class=\"medius-aisummary\" data-starttime=\"00:03:07\">00:03:07</a>). Each plays a critical role in ensuring functionality, data access, and privacy compliance — especially in markets like Germany, where data governance is strict. A reflection on earlier Azure AI work illustrates why previous solutions failed to gain production trust — users saw them as “black boxes” without transparency. The talk ties this lesson to the importance of enabling compliance oversight in current implementations via Microsoft 365 Copilot and the underlying MCP (Microsoft Copilot Platform) architecture (<a class=\"medius-aisummary\" data-starttime=\"00:03:59\">00:03:59</a>).</p>\n\n<p><strong>Demonstration: Connecting Agents Through Worker Queue and Foundry:</strong> Around this midpoint, a live demonstration begins showing Worker Queue’s integration with Foundry and Fabric IQ (<a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>). The speaker illustrates how an AI agent interacts with real user data — such as summarizing emails — through secure approval steps (“human in the loop”) ensuring user consent when connecting to Work Mail. Visual output from Copilot Studio demonstrates initialization of data agents, querying of mails, and the retrieval of contextual summaries (<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>). Using multiple connectors like Copilot, Teams, or databases through MCP servers, the demonstration underscores how cross-platform orchestration occurs securely and interactively. The result showcases agents merging Worker Queue and Foundry IQ data to automatically generate summaries and draft communications.</p>\n\n<p><strong>Fabric Ontologies, Knowledge Graphs, and Semantic Context:</strong> The discussion then transitions into Fabric IQ’s ontology and how it enables agents to understand relationships between data points in graph form instead of traditional SQL lookups (<a class=\"medius-aisummary\" data-starttime=\"00:13:19\">00:13:19</a>). Using examples from enterprise cases such as Swiss Airlines, the speaker explains the bronze-silver-gold data enrichment hierarchy feeding semantic models. Ontologies define nodes and relations — for instance, product-to-store performance data — making it easier for agents to infer context quickly. This interconnected design improves agent responsiveness and contextual reasoning, and can be published to Microsoft 365 Copilot for direct conversational querying and visualization of data, including plotting charts directly within M365 (<a class=\"medius-aisummary\" data-starttime=\"00:18:23\">00:18:23</a>).</p>\n\n<p><strong>Advanced Implementation and Evaluation:</strong> The closing part focuses on advanced agent integration using SDKs and validation workflows (<a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>). Developers can configure Fabric data agents, connect them to evaluation notebooks, and run accuracy testing before deployment. The talk emphasizes continuous monitoring through Microsoft Purview, enabling visibility into sensitive data interactions and user behavior within Agent 365 (<a class=\"medius-aisummary\" data-starttime=\"00:26:52\">00:26:52</a>). By combining transparency and governance, agents now operate within a secure, inspectable framework rather than the opaque “black box” systems of the past. The speaker concludes by thanking attendees and encouraging them to explore released notebooks and repositories post-conference (<a class=\"medius-aisummary\" data-starttime=\"00:27:42\">00:27:42</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"19:20 - 19:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:45:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/81eb53e4-8657-49d5-b228-c473ded735dc","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/81eb53e4-8657-49d5-b228-c473ded735dc","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/81eb53e4-8657-49d5-b228-c473ded735dc","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/81eb53e4-8657-49d5-b228-c473ded735dc","speakerIds":["1768245294609build26-1760032715730001qqSb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.903Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/81eb53e4-8657-49d5-b228-c473ded735dc","ogTitle":"Building a Multi-Agent Workflow in Microsoft Fabric","ogDescription":"Learn how Microsoft Fabric and the 3 IQs, Work IQ, Fabric IQ, and Foundry IQ, enable dynamic, context‑aware multi‑agent workflows. See how semantic models, lakehouses, and pipelines form agent‑ready foundations, and how specialized agents collaborate across planning, retrieval, summarization, and execution. Discover how natural language, automation, and orchestration power real multi‑agent scenarios in Microsoft Fabric.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1760032715730001qqSb","RegistrantKey":"1768245294609build26-1760032715730001qqSb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"c65807f6-36f9-4585-aa7e-28b9a815df0d","startDateTime":"2026-06-02T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prashant G Bhoyar, Rie Merritt, Mitchel Sellers, Stephen Simon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774050746002001lFvK","langLocale":"en-US","sessionId":"1774050746002001lFvK","sessionInstanceId":"1774050746002001lFvK","sessionCode":"DEM377","sessionCodeNormalized":"EN-US-DEM377","title":"Building Impact Through Developer Communities","sortTitle":"building impact through developer communities","description":"Join community leaders from.NET, Azure, Fabric and Global AI, to explore what it takes to build and grow tech communities. Hear their journeys, lessons learned, and how they stepped into leadership. Learn how to get involved, start a community, and connect with others who share your passion for technology.","aiDescription":"<p><strong>Introduction and Panel Overview:</strong> The session opens with host Gallima Basa greeting viewers and confirming audio before introducing herself as the Azure Community Lead, explaining her work to empower communities globally (<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>). She outlines that the discussion will feature leaders representing .NET, Azure, Fabric, and Global AI communities to explore what it means to build and sustain technical communities, how to participate in them, and the leadership lessons gained along the way (<a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>). The format encourages audience engagement as the conversation moves into introductions from the panelists.</p>\n\n<p><strong>Community Leader Introductions:</strong> Mitchell Sellers begins the introductions, sharing his journey within the .NET community since 2008 and his path into open-source and conference leadership through the .NET Foundation (<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:37\">00:01:37</a>). Next, Prashanji Boyer from Washington, DC recounts his experience evolving from attendee to organizer, running user groups and regional Microsoft tech conferences (<a class=\"medius-aisummary\" data-starttime=\"00:01:42\">00:01:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:14\">00:02:14</a>). Rhee Merritt follows, describing her Microsoft role managing Azure Data influencer communities and events like Fabric and SQL Con, while also running a local user group in Atlanta (<a class=\"medius-aisummary\" data-starttime=\"00:02:15\">00:02:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>). Each provides insight into how professional and volunteer roles merge to sustain grassroots engagement.</p>\n\n<p><strong>Motivations Behind Leadership:</strong> When asked about their motivation to step into leadership, Rhee candidly admits her initial drive stemmed from wanting to ensure things were done right but grew into a passion for giving back after realizing how community involvement shaped her career advancement (<a class=\"medius-aisummary\" data-starttime=\"00:03:13\">00:03:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:13\">00:04:13</a>). Prashanji credits his mentors’ encouragement and the emerging opportunity in AI-related technologies around 2017, which prompted him to create local meetups and collaborate with global groups to fill a knowledge gap (<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:39\">00:05:39</a>). Mitchell echoes a similar sentiment, emphasizing that leadership enables professionals to mentor newcomers while reinforcing their own learning, thus continuing the community’s cycle of growth (<a class=\"medius-aisummary\" data-starttime=\"00:05:45\">00:05:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:34\">00:06:34</a>).</p>\n\n<p><strong>How to Get Involved and Start Small:</strong> When discussing how newcomers can join or contribute, Rhee highlights that simply attending events already makes one part of the community, and deeper involvement should start small — volunteering for tasks such as registration or communicating event details (<a class=\"medius-aisummary\" data-starttime=\"00:06:35\">00:06:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:58\">00:07:58</a>). Prashanji advises identifying a personal “sweet spot,” like helping with marketing, social media, or logistics, while Mitchell recommends networking at after-hours events and conferences where organizers and creators naturally connect (<a class=\"medius-aisummary\" data-starttime=\"00:08:18\">00:08:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>). They collectively reinforce that consistent, genuine contributions in small acts build trust and open doors to greater roles over time.</p>\n\n<p><strong>Building and Running User Groups:</strong> The panel next turns to the topic of creating a user group. Rhee details the application process for Microsoft-recognized communities, which requires focusing at least half of sessions on Microsoft technologies, holding regular meetings, and maintaining at least two leaders for continuity (<a class=\"medius-aisummary\" data-starttime=\"00:10:24\">00:10:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>). She cautions that starting from scratch is difficult and encourages beginners to first volunteer with established groups to learn about securing venues, sponsors, and speakers. The main takeaway is to grow gradually with mentorship and avoid overcommitting before understanding the full operational workload (<a class=\"medius-aisummary\" data-starttime=\"00:11:46\">00:11:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:41\">00:12:41</a>).</p>\n\n<p><strong>Memorable Successes in Community Impact:</strong> The leaders then share personal success stories. Prashanji recalls a participant who transitioned to an IT career after attending his local workshops, attributing his progress to hands-on exposure through community labs (<a class=\"medius-aisummary\" data-starttime=\"00:12:58\">00:12:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:07\">00:14:07</a>). Mitchell treasures seeing first-time speakers grow into confident professionals who later lead their own events or achieve MVP recognition (<a class=\"medius-aisummary\" data-starttime=\"00:14:16\">00:14:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:22\">00:15:22</a>). Rhee shares how one small act of kindness — welcoming a newcomer at a user group — transformed that attendee’s career path, illustrating the unseen power of inclusivity and personal connection (<a class=\"medius-aisummary\" data-starttime=\"00:15:31\">00:15:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:34\">00:17:34</a>).</p>\n\n<p><strong>Lessons from Events and Closing Thoughts:</strong> In discussing recent event lessons, Mitchell reflects on a post-pandemic in-person conference that underperformed due to travel costs, stressing the importance of understanding audience capabilities and trusting instincts when making planning decisions (<a class=\"medius-aisummary\" data-starttime=\"00:17:56\">00:17:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:46\">00:19:46</a>). Prashanji shares a pragmatic tip — charging a small event fee greatly improves attendance reliability compared to free events (<a class=\"medius-aisummary\" data-starttime=\"00:19:54\">00:19:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:56\">00:20:56</a>). Rhee concludes by describing Fabric Con’s emphasis on fostering community-driven spaces, such as lounges and networking areas free from sales pitches, reinforcing that conferences succeed when attendees feel connected, welcomed, and inspired to grow their networks (<a class=\"medius-aisummary\" data-starttime=\"00:21:06\">00:21:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:35\">00:23:35</a>). The session ends with applause for the panelists, wrapping up the theme that community, at its heart, thrives through authentic human interaction and shared learning.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:30 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/eb5c4591-f445-4c81-8428-6293a5ea8192","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/eb5c4591-f445-4c81-8428-6293a5ea8192","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/eb5c4591-f445-4c81-8428-6293a5ea8192","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/eb5c4591-f445-4c81-8428-6293a5ea8192","speakerIds":["1768245294609build26-1771723064928001lcoH","1768245294609build26-1778277011891001WLB9","1768245294609build26-1777917053803001zEGS","1768245294609build26-1774323608465001D4tS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.632Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/eb5c4591-f445-4c81-8428-6293a5ea8192","ogTitle":"Building Impact Through Developer Communities","ogDescription":"Join community leaders from.NET, Azure, Fabric and Global AI, to explore what it takes to build and grow tech communities. Hear their journeys, lessons learned, and how they stepped into leadership. Learn how to get involved, start a community, and connect with others who share your passion for technology.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778277011891001WLB9","RegistrantKey":"1768245294609build26-1778277011891001WLB9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774323608465001D4tS","RegistrantKey":"1768245294609build26-1774323608465001D4tS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771723064928001lcoH","RegistrantKey":"1768245294609build26-1771723064928001lcoH"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777917053803001zEGS","RegistrantKey":"1768245294609build26-1777917053803001zEGS"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"22c2363c-face-4935-8ff2-d850d5f47d3f","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Edo Segal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776886872858001iZkX","langLocale":"en-US","sessionId":"1776886872858001iZkX","sessionInstanceId":"1776886872858001iZkX","sessionCode":"DEMSP390","sessionCodeNormalized":"EN-US-DEMSP390","title":"Create multimodal AI agents with persistent memory","sortTitle":"create multimodal ai agents with persistent memory","description":"Your users reach you through your website, your app, your store, your support line. And whether they were in touch five minutes or five days ago, every channel starts from zero, every time. Multimodal AI agents with persistent memory close that gap, and you can have them up and running now. In this session, Napster CTPO Edo Segal builds a working video AI agent live using the Napster Omniagent API. You'll leave with the architecture, the code patterns, and a clear next step to build your own agent that lives across every surface your organization touches.","aiDescription":"<p><strong>Introduction and Outlook:</strong> The session opens with a welcome to Build 2027 <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, highlighting how exciting this year’s event is compared to prior builds. The speaker emphasizes the breadth of Microsoft’s innovation across hardware and cloud technology <a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>, and sets forward-looking expectations for next year’s conference, predicting that attendees will interact with AI “agents” that behave like people, not just chatbots. The mood is visionary as the audience is invited to imagine advanced, lifelike digital assistants integrated into daily experiences.</p>\n\n<p><strong>Napster’s AI Vision and Foundry Integration:</strong> The presentation transitions to an introduction from Napster <a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>, where the team showcases a “crew of AI agents” that help users accomplish their tasks <a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>. These agents do not physically exist but are fully digital and AI-driven. The speaker explains how Foundry enables creation of multimodal AI personas—video, voice, and text—through a few clicks. Napster’s product stack includes hardware solutions such as “The View,” a holographic display that extends user interaction, and kiosks suitable for busy public environments <a class=\"medius-aisummary\" data-starttime=\"00:01:38\">00:01:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:56\">00:01:56</a>. Attendees are urged to visit the booth to see these experiences firsthand.</p>\n\n<p><strong>Omni Agent API and Microsoft Azure Collaboration:</strong> The next segment dives into the technical foundation—the Omni Agent API <a class=\"medius-aisummary\" data-starttime=\"00:03:38\">00:03:38</a>, which unifies modalities (text, video, phone, chat) under one agent able to persist memory and maintain relationships with users—creating a new paradigm of “human emulation” <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:16\">00:04:16</a>. This innovation builds relational intelligence between humans and their AI counterparts. Later, Microsoft’s representative joins to announce Napster’s public preview as a native Azure offering <a class=\"medius-aisummary\" data-starttime=\"00:04:38\">00:04:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>. The integration offers unified billing through Azure Marketplace, easy provisioning via the Azure portal, and seamless linkage between Napster’s experiential layer and Foundry’s intelligence layer—streamlining enterprise-level deployment.</p>\n\n<p><strong>Live Demo and Developer Workflow:</strong> The speaker then demonstrates development of an Omni Agent <a class=\"medius-aisummary\" data-starttime=\"00:06:03\">00:06:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:06\">00:07:06</a>, showing how a developer can simply add prompts and API keys to create an intelligent assistant that understands site code directly from a Git repository. The process automates everything locally via an MCP server embedded in JavaScript—what they call “Edge MCP”—enabling local cognition in real time <a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:09\">00:08:09</a>. This allows agents to perform tasks directly through the DOM without external calls. Igor and others elaborate how AI models like Opus 4.8 make it possible to predict user behavior across a full code base and automatically create Azure-hosted visual agents within minutes <a class=\"medius-aisummary\" data-starttime=\"00:09:12\">00:09:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:38\">00:10:38</a>.</p>\n\n<p><strong>Engineering Innovations and Real-World Applications:</strong> The presentation then highlights Napster’s breakthrough in reducing video-agent costs by 20x—down to one cent per minute—enabling widespread deployment of video avatars on websites <a class=\"medius-aisummary\" data-starttime=\"00:11:18\">00:11:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>. The discussion reaffirms a shift toward human-like interfaces where users interact through conversation rather than code. Technical diagrams show how apps, agents, and Edge MCP servers form a three-layer system—the app (truth), the bridge (action), and the agent (persona)—leading to seamless browser control and state awareness <a class=\"medius-aisummary\" data-starttime=\"00:13:06\">00:13:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. This enables AI coworkers capable of performing actual business functions—from helping customers in retail stores to assisting patients at hospitals—all using the same underlying framework <a class=\"medius-aisummary\" data-starttime=\"00:15:18\">00:15:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:58\">00:15:58</a>.</p>\n\n<p><strong>Demo Wrap-Up and Closing Remarks:</strong> To conclude, the team shows a real-time demo query—searching for OLED TVs under $2000—illustrating how the Edge MCP server enables dynamic interactions and context awareness <a class=\"medius-aisummary\" data-starttime=\"00:16:40\">00:16:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:20\">00:18:20</a>. Participants see live capability calls and state updates of an AI-agent browser control system operating autonomously. The session ends with an invitation to scan QR codes for tokens, prompts, and participation in a hackathon with prizes <a class=\"medius-aisummary\" data-starttime=\"00:19:02\">00:19:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:40\">00:19:40</a>. The speakers thank partners Microsoft, product managers, and engineers for enabling this next-generation AI experience, signaling the dawn of practical, multimodal digital coworkers.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:00 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9228bcee-53b2-4fd0-b6a7-83ca65070cfd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9228bcee-53b2-4fd0-b6a7-83ca65070cfd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9228bcee-53b2-4fd0-b6a7-83ca65070cfd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9228bcee-53b2-4fd0-b6a7-83ca65070cfd","speakerIds":["1768245294609build26-1776448795539001JeI6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.385Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Napster"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9228bcee-53b2-4fd0-b6a7-83ca65070cfd","ogTitle":"Create multimodal AI agents with persistent memory","ogDescription":"Your users reach you through your website, your app, your store, your support line. And whether they were in touch five minutes or five days ago, every channel starts from zero, every time. Multimodal AI agents with persistent memory close that gap, and you can have them up and running now. In this session, Napster CTPO Edo Segal builds a working video AI agent live using the Napster Omniagent API. You'll leave with the architecture, the code patterns, and a clear next step to build your own agent that lives across every surface your organization touches.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776448795539001JeI6","RegistrantKey":"1768245294609build26-1776448795539001JeI6"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Live%20Demo_1779385317341001sOfV_1780513463412001YxSh.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"75fc190c-dba2-48c8-8f79-c8d7e04d055e","startDateTime":"2026-06-03T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prashant G Bhoyar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290489938001Zds8","langLocale":"en-US","sessionId":"1776290489938001Zds8","sessionInstanceId":"1776290489938001Zds8","sessionCode":"DEM368","sessionCodeNormalized":"EN-US-DEM368","title":"Data Science & Machine Learning with Microsoft Fabric","sortTitle":"data science & machine learning with microsoft fabric","description":"Explore Microsoft Fabric’s unified Data Science experience in this fast-paced 25-minute demo. See how to ingest and prepare data with OneLake and Lakehouse, build models using Notebooks, Spark, and SynapseML, and track experiments with MLflow. Learn how to operationalize models with batch scoring and integrate results into Power BI. The session also showcases building a simple generative AI Q&A solution using Fabric Data Agents, giving you a practical end-to-end view in a short time.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:30 - 21:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c9ea1330-003f-4964-8ae8-28490b0d3eb2","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c9ea1330-003f-4964-8ae8-28490b0d3eb2","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c9ea1330-003f-4964-8ae8-28490b0d3eb2","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c9ea1330-003f-4964-8ae8-28490b0d3eb2","speakerIds":["1768245294609build26-1771723064928001lcoH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.887Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c9ea1330-003f-4964-8ae8-28490b0d3eb2","ogTitle":"Data Science & Machine Learning with Microsoft Fabric","ogDescription":"Explore Microsoft Fabric’s unified Data Science experience in this fast-paced 25-minute demo. See how to ingest and prepare data with OneLake and Lakehouse, build models using Notebooks, Spark, and SynapseML, and track experiments with MLflow. Learn how to operationalize models with batch scoring and integrate results into Power BI. The session also showcases building a simple generative AI Q&A solution using Fabric Data Agents, giving you a practical end-to-end view in a short time.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771723064928001lcoH","RegistrantKey":"1768245294609build26-1771723064928001lcoH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"e8e6f8f9-6351-4663-a2d4-a71d068a0458","startDateTime":"2026-06-04T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prashant G Bhoyar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779814066720001okQD","langLocale":"en-US","sessionId":"1779814066720001okQD","sessionInstanceId":"1779814066720001okQD","sessionCode":"DEM368-R1","sessionCodeNormalized":"EN-US-DEM368-R1","title":"Data Science & Machine Learning with Microsoft Fabric","sortTitle":"data science & machine learning with microsoft fabric","description":"Explore Microsoft Fabric’s unified Data Science experience in this fast-paced 25-minute demo. See how to ingest and prepare data with OneLake and Lakehouse, build models using Notebooks, Spark, and SynapseML, and track experiments with MLflow. Learn how to operationalize models with batch scoring and integrate results into Power BI. The session also showcases building a simple generative AI Q&A solution using Fabric Data Agents, giving you a practical end-to-end view in a short time.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda Overview:</strong> The video begins with the speaker explaining that machine learning (ML) within Microsoft Fabric remains one of its least explored capabilities (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>). The session’s agenda includes discussing why traditional ML is still significant, even amidst the rise of generative AI. The speaker outlines when and why teams may prefer classic ML models—such as when low latency, computational efficiency, consistent results, and cost control are essential (<a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>). He highlights that while multiple Microsoft and open-source tools can run ML models, the focus of the talk is on how these workloads operate directly within Microsoft Fabric, which serves as a unified data and AI platform integrating services like Data Factory, Real-Time Intelligence, and Power BI (<a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>).</p>\n\n<p><strong>Overview of Microsoft Fabric Capabilities:</strong> The speaker provides an overview of Fabric, describing it as a software-as-a-service (SaaS) offering that unifies analytics, data warehousing, engineering, and business intelligence within a single environment (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>). He explains that different roles use different Fabric components—data engineers focus on Data Factory and Warehouses, analytics engineers on Power BI, and data scientists on integrating Azure Machine Learning with Fabric. The session introduces a standard data science lifecycle: defining a problem, gathering and preparing data, conducting experimentation, training models, and generating insights (<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>). This process integrates seamlessly with Fabric components like the Lakehouse for data storage, Data Wrangler for transformation, and notebooks for model training and tracking via MLFlow. The speaker notes that resulting models can then be deployed within Fabric and consumed by Power BI reports or other custom applications (<a class=\"medius-aisummary\" data-starttime=\"00:03:47\">00:03:47</a>).</p>\n\n<p><strong>Machine Learning Demonstration in Fabric:</strong> Transitioning to a live demo, the speaker guides viewers through a scenario involving a public healthcare dataset used to build a predictive model for diabetes (<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>). He demonstrates creating a notebook within a Fabric workspace called “Labs” and loading the dataset stored in Azure Blob Storage into a dataframe (<a class=\"medius-aisummary\" data-starttime=\"00:06:15\">00:06:15</a>). The data is converted into a pandas dataframe, split into training (70%) and validation (30%) sets (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>). Two models are trained—a linear regression and a decision tree regressor—using PySpark within the Fabric notebook environment. The experiments are logged, compared, and tracked within MLFlow (<a class=\"medius-aisummary\" data-starttime=\"00:10:10\">00:10:10</a>). After evaluating performance metrics, the decision tree regressor is found to perform better, illustrating the model selection process based on defined evaluation criteria.</p>\n\n<p><strong>Model Deployment and Integration:</strong> Once the optimal experiment is chosen, the speaker demonstrates saving it as a deployable ML model within Fabric (<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>). The model settings can be customized with descriptions and versioning. He then discusses how models can be consumed—via model endpoints exposed through Fabric—to integrate predictions into other Microsoft applications. Compared with Azure Machine Learning in Foundry, Fabric’s key advantages are its simplicity, unified environment, and reduced management overhead because it handles all services in a SaaS-based architecture (<a class=\"medius-aisummary\" data-starttime=\"00:14:20\">00:14:20</a>). This integration ensures users can deploy, manage, and consume ML models entirely within Fabric’s ecosystem, supporting smoother collaboration between data scientists and business users.</p>\n\n<p><strong>Introducing Fabric Data Agents:</strong> The next section shifts focus to Fabric’s data agent capabilities—a generative AI-powered feature that lets everyday users query their data intuitively without complex SQL writing (<a class=\"medius-aisummary\" data-starttime=\"00:14:49\">00:14:49</a>). Unlike notebooks designed for data scientists, data agents democratize insight generation. The speaker shows how to create a data agent by selecting “New Item” in Fabric, naming the agent, and connecting it to existing data sources like lakehouses or warehouses (<a class=\"medius-aisummary\" data-starttime=\"00:16:30\">00:16:30</a>). He walks through an example using sales data to identify top-performing products. The agent automatically converts natural language queries into SQL statements, executes them, retrieves results, and summarizes the findings (<a class=\"medius-aisummary\" data-starttime=\"00:18:13\">00:18:13</a>). This automation reduces dependency on technical teams and expedites access to business insights.</p>\n\n<p><strong>Integration Across Microsoft Ecosystem and Conclusion:</strong> The final segment shows how Fabric data agents can be deployed across the Microsoft ecosystem (<a class=\"medius-aisummary\" data-starttime=\"00:20:06\">00:20:06</a>). Agents can connect to Copilot Studio, integrate within Microsoft 365 Copilot, or be exposed as MCP servers for external use (<a class=\"medius-aisummary\" data-starttime=\"00:21:29\">00:21:29</a>). This interoperability enables organizations to scale intelligence consistently across tools and teams. The speaker concludes by reaffirming that both traditional ML and generative AI capabilities coexist synergistically within Fabric. He encourages viewers to reach out with follow-up questions and closes the session with a thank-you message, offering his LinkedIn contact via QR code (<a class=\"medius-aisummary\" data-starttime=\"00:23:16\">00:23:16</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4924e612-6605-49b7-9f01-7b9fab96e6e7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4924e612-6605-49b7-9f01-7b9fab96e6e7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4924e612-6605-49b7-9f01-7b9fab96e6e7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4924e612-6605-49b7-9f01-7b9fab96e6e7","speakerIds":["1768245294609build26-1771723064928001lcoH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.239Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4924e612-6605-49b7-9f01-7b9fab96e6e7","ogTitle":"Data Science & Machine Learning with Microsoft Fabric","ogDescription":"Explore Microsoft Fabric’s unified Data Science experience in this fast-paced 25-minute demo. See how to ingest and prepare data with OneLake and Lakehouse, build models using Notebooks, Spark, and SynapseML, and track experiments with MLflow. Learn how to operationalize models with batch scoring and integrate results into Power BI. The session also showcases building a simple generative AI Q&A solution using Fabric Data Agents, giving you a practical end-to-end view in a short time.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771723064928001lcoH","RegistrantKey":"1768245294609build26-1771723064928001lcoH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"4a98f1b9-9cec-4a55-a6c0-0e95488bf746","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Adrian Macias","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775240993291001janl","langLocale":"en-US","sessionId":"1775240993291001janl","sessionInstanceId":"1775240993291001janl","sessionCode":"DEMSP386","sessionCodeNormalized":"EN-US-DEMSP386","title":"Develop faster on Windows with AI playbooks and local agents","sortTitle":"develop faster on windows with ai playbooks and local agents","description":"In this session you'll experience a practical path for a streamlined setup-to-ship workflow. Learn how you can use innovations in local coding on AMD's Ryzen™ AI PCs to optimize workflows powered by AI Playbooks to reduce friction and accelerate the path from idea to first commit. Finish with a clear example showing how to build once, run locally, and scale across Windows systems optimized for AMD hardware. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The speaker opens the session by explaining that there is a large amount of new information and features to cover in the demo <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>. They invite participants to ask questions at any time for an interactive discussion. As a member of AMD's Developer Acceleration Team, the speaker focuses on the client side of AI—highlighting AI PCs that merge local compute with cloud processing. The team collaborates closely with ISVs and OEMs to accelerate AI app deployment, supported by experts across full-stack and backend implementations <a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>. The session centers around introducing AMD’s “Playbooks”—developer resources designed to streamline application creation, alongside foundational technologies like Lemonade and Lemonade Routers within the ROCm stack <a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>.</p>\n\n<p><strong>AI Use Cases and Developer Empowerment:</strong> The discussion shifts to AI use cases that have evolved in recent years, such as workflow automation, knowledge bases, gaming, and coding as a ubiquitous foundation <a class=\"medius-aisummary\" data-starttime=\"00:02:20\">00:02:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. Participants are invited to a hands-on workshop about coding with agents. The speaker emphasizes “choice” as a pivotal theme across modern AI development <a class=\"medius-aisummary\" data-starttime=\"00:03:29\">00:03:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>. They describe layers of choice in model quality, cost, locality (CPU/GPU/NPU or client-cloud split), and modality (text, vision, image, audio) <a class=\"medius-aisummary\" data-starttime=\"00:03:52\">00:03:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:06\">00:05:06</a>. These choices empower developers to define where and how their AI workloads run, dynamically balancing user experience, cost efficiency, and computing logistics. The emerging omnimodal AI models combine text, speech, and vision—though the standardization of serving such models remains a work in progress <a class=\"medius-aisummary\" data-starttime=\"00:05:12\">00:05:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>.</p>\n\n<p><strong>Agentic Coding and Dynamic Routing Choices:</strong> The speaker delves into examples of agentic coding—demonstrating how developers can set agents for specific tasks and dynamically choose computation modes <a class=\"medius-aisummary\" data-starttime=\"00:06:04\">00:06:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>. For heavy architectural tasks, developers might opt for larger cloud models, while quicker debugging or creative iterations can run locally—even without Wi-Fi connection. The demo showcases dynamic model routing based on resource availability and workflow context <a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>. Next, the speaker introduces Lemonade, part of the ROCm stack, which abstracts backend complexity by automatically routing workloads across different processors through semantic rules <a class=\"medius-aisummary\" data-starttime=\"00:08:00\">00:08:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:19\">00:09:19</a>. Lemonade integrates easily with frameworks such as Codex, Claude, or Gaia, and selects locally optimized models automatically, making intelligent routing seamless <a class=\"medius-aisummary\" data-starttime=\"00:09:25\">00:09:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:05\">00:10:05</a>. Through a weather-dashboard demo <a class=\"medius-aisummary\" data-starttime=\"00:10:10\">00:10:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:17\">00:11:17</a>, the session illustrates how Lemonade supports creative, agile workflows locally and dynamically shifts resources when needed.</p>\n\n<p><strong>Playbooks and Omni-Modal Demonstrations:</strong> AMD’s “Playbooks” concept is detailed as a unified open-source set of guides for developers across skill levels—from kernel optimization to image generation UI integration <a class=\"medius-aisummary\" data-starttime=\"00:11:51\">00:11:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>. The collection offers full-stack recipes adaptable for Rock’em kernels, Lemonade-powered abstraction, and cross-platform AI app development. These resources recently launched for Radeon GPUs and client devices, with expansion toward Halo appliance PCs and EPIC server platforms. The speaker highlights the open GitHub repository where these playbooks are hosted, encouraging developers to explore unreleased clustering and multimodal capabilities <a class=\"medius-aisummary\" data-starttime=\"00:13:33\">00:13:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:10\">00:14:10</a>. The demo on “omni-modal routing” shows Lemonade’s ability to combine models intelligently—such as linking image generation, narration, and editing functions—to enable tasks like generating and narrating interactive stories <a class=\"medius-aisummary\" data-starttime=\"00:15:53\">00:15:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:30\">00:17:30</a>. This integration demonstrates Lemonade’s ability to manage complex multimodal queries contextually without requiring deep technical knowledge.</p>\n\n<p><strong>Semantic Filtering, Cost Optimization, and Developer Ecosystem:</strong> Further examples show Lemonade and VLLM router being used for semantic filtering and secure routing <a class=\"medius-aisummary\" data-starttime=\"00:18:42\">00:18:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>. The system dynamically distinguishes private data to route locally or expensive cloud-based tasks to appropriate endpoints, optimizing both cost and privacy. For instance, personal queries with sensitive data remain on-device, while trivial queries or small jobs run locally to reduce cloud costs <a class=\"medius-aisummary\" data-starttime=\"00:19:15\">00:19:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:50\">00:20:50</a>. The backend “cortex” architecture unifies diverse device APIs such as Vulkan, Stable Diffusion, Whisper, and ROCm under a standard inference API inspired by OpenAI’s endpoint format <a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:17\">00:22:17</a>. Together with model management and application-level utility layers, this approach democratizes AI development through the open-source developer community <a class=\"medius-aisummary\" data-starttime=\"00:22:27\">00:22:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:03\">00:23:03</a>. The speaker highlights developer enthusiasm—mentioning community projects and cross-vendor contributions—and reveals the upcoming “Halo box,” a headless PC appliance designed for continuous AI agent operation and background automation tasks <a class=\"medius-aisummary\" data-starttime=\"00:23:26\">00:23:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:22\">00:24:22</a>.</p>\n\n<p><strong>Conclusion and Call to Action:</strong> The session closes with an invitation to explore AMD’s Developer Portal (<a class=\"medius-aisummary\" data-starttime=\"00:24:26\">00:24:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:26\">00:25:26</a>), where visitors will find announcements on Halo, Lemonade updates, and Playbook workshops. The speaker underscores AMD’s emphasis on hands-on, global developer engagement—with recent workshops drawing thousands of attendees worldwide. They encourage new developers to register for updates and access cloud compute credits for AMD Instinct-class devices. After addressing final audience questions about hardware pricing and launch timing <a class=\"medius-aisummary\" data-starttime=\"00:25:33\">00:25:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:26:31\">00:26:31</a>, the presenter thanks attendees and concludes the interactive session, emphasizing AMD’s commitment to accessible, scalable AI innovation through open tools and community collaboration.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:30 - 17:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/3e9502fa-d5ca-49a3-a3bb-69d4be31d73f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/3e9502fa-d5ca-49a3-a3bb-69d4be31d73f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/3e9502fa-d5ca-49a3-a3bb-69d4be31d73f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/3e9502fa-d5ca-49a3-a3bb-69d4be31d73f","speakerIds":["1768245294609build26-1776355971597001QUS8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.893Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/3e9502fa-d5ca-49a3-a3bb-69d4be31d73f","ogTitle":"Develop faster on Windows with AI playbooks and local agents","ogDescription":"In this session you'll experience a practical path for a streamlined setup-to-ship workflow. Learn how you can use innovations in local coding on AMD's Ryzen™ AI PCs to optimize workflows powered by AI Playbooks to reduce friction and accelerate the path from idea to first commit. Finish with a clear example showing how to build once, run locally, and scale across Windows systems optimized for AMD hardware. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776355971597001QUS8","RegistrantKey":"1768245294609build26-1776355971597001QUS8"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Build_2026_Adrian_Breakout_June_2_v3_1780273438041001PabO_1780514838357001lShA.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Windows SDKs","logicalValue":"Windows SDKs"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Windows Development","logicalValue":"Windows Development"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"a91ea363-1226-482c-90bf-ab133ad3ebde","startDateTime":"2026-06-03T19:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christine Flora","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290078162001i6A7","langLocale":"en-US","sessionId":"1776290078162001i6A7","sessionInstanceId":"1776290078162001i6A7","sessionCode":"DEM360","sessionCodeNormalized":"EN-US-DEM360","title":"Discovering Power App Data with M365 Copilot using MCP","sortTitle":"discovering power app data with m365 copilot using mcp","description":"With millions of users world-wide, Power Apps and Dataverse now sit atop critical business data, but developers face challenges with discoverability and schema understanding. This session introduces Power Apps Model Context Protocol (MCP), which empowers M365 Copilot to access app metadata, Dataverse schemas, relationships, and permissions. Learn how MCP enables Copilot to answer natural-language queries, aggregate data, and deliver real-time insights, plus best practices for Copilot-ready apps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Goals:</strong> Christine Flora introduces the DEM 360 session and explains her background as a Microsoft MVP focused on Power Apps and Power Platform <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>-<a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>. She outlines the agenda, including an overview of the new Power Apps MCP tool, why MCP matters for securely exposing business data across Microsoft 365 applications like Outlook, Excel, Word, Teams, Copilot, and Copilot Studio, and how to create custom tools and declarative agents. She also previews a walkthrough of enabling MCP support in Power Apps and explains that prerequisites and deployment guidance will be covered at the end.</p>\n\n<p><strong>Why MCP Matters for Power Apps:</strong> The speaker explains that MCPs act as a \"universal translator\" for agentic AI experiences, enabling applications and copilots to securely interact with enterprise data <a class=\"medius-aisummary\" data-starttime=\"00:01:24\">00:01:24</a>-<a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>. She describes Power Platform as a low-code and pro-code ecosystem that empowers both business users and developers to create applications, automate workflows, analyze data, and build AI-driven solutions using connected data sources. The session highlights integrations with tools such as VS Code, GitHub Copilot, Claude, DevOps, and Copilot Studio. Christine also notes that the demo will move between Power Apps Maker Portal, VS Code, Claude, and Microsoft 365 to show MCP capabilities in action.</p>\n\n<p><strong>Live Demo of MCP and Declarative Agents:</strong> Christine demonstrates an equipment maintenance Power App that tracks inspections, work orders, and equipment history <a class=\"medius-aisummary\" data-starttime=\"00:03:41\">00:03:41</a>-<a class=\"medius-aisummary\" data-starttime=\"00:08:43\">00:08:43</a>. Initially, a standard Copilot query cannot access app-specific maintenance data because it only searches Microsoft 365 content. After enabling MCP and deploying a declarative agent, Copilot can retrieve live application data directly inside Microsoft 365 using natural language prompts. She shows interactive work order lists, record drilldowns, sorting, and direct links into the Power App. The demo then expands into richer custom MCP tools, including KPI dashboards, equipment cards, issue tracking, and a visual work-order timeline that supports filtering, hover details, scheduling analysis, and live operational insights directly inside Copilot, Word, Outlook, and Excel.</p>\n\n<p><strong>Creating and Configuring Custom MCP Tools:</strong> The walkthrough then shifts into the technical implementation process <a class=\"medius-aisummary\" data-starttime=\"00:08:47\">00:08:47</a>-<a class=\"medius-aisummary\" data-starttime=\"00:15:31\">00:15:31</a>. In the Power Apps Maker Portal, Christine enables the preview Copilot control and MCP support for the app, exposing built-in CRUD capabilities and custom tools. She creates a new custom tool by defining prompts, selecting AI models, configuring Dataverse tables and fields, and generating a JSON payload. Using VS Code with Power Platform tools and Claude Code Pro, she imports the generated JSON and uses MCP skills to automatically create Fluent UI components for the custom experience. After previewing and refining the generated UI, she pastes the Fluent UI code back into Power Apps, publishes the app, and packages the declarative agent for upload into Microsoft Teams and Microsoft 365 app catalogs.</p>\n\n<p><strong>Prerequisites, Deployment, and Closing Notes:</strong> The final section covers licensing, permissions, and deployment requirements for using Power Apps MCP features <a class=\"medius-aisummary\" data-starttime=\"00:15:31\">00:15:31</a>-<a class=\"medius-aisummary\" data-starttime=\"00:21:12\">00:21:12</a>. Users and developers need Power Apps licenses, Copilot licenses, and proper data permissions. Developers also need customizer or admin privileges, preview Maker Portal access, VS Code, Node.js, Power Platform tools, and either GitHub Copilot Pro or Claude Code Pro for skill generation. Christine explains that MCP support is currently preview-only and mainly available for model-driven apps, though broader support is planned. She concludes by answering audience questions about government cloud availability, cloud flow triggers, external data connectors, and virtual tables, emphasizing that MCP interactions behave similarly to native Power Apps data operations. The session ends with links to documentation, GitHub repositories, and additional resources for continued experimentation.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"19:50 - 20:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:15:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/47a31e54-7ea0-4b32-be8e-f7c6400735e6","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/47a31e54-7ea0-4b32-be8e-f7c6400735e6","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/47a31e54-7ea0-4b32-be8e-f7c6400735e6","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/47a31e54-7ea0-4b32-be8e-f7c6400735e6","speakerIds":["1768245294609build26-1772578847951001DH4K"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.904Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/47a31e54-7ea0-4b32-be8e-f7c6400735e6","ogTitle":"Discovering Power App Data with M365 Copilot using MCP","ogDescription":"With millions of users world-wide, Power Apps and Dataverse now sit atop critical business data, but developers face challenges with discoverability and schema understanding. This session introduces Power Apps Model Context Protocol (MCP), which empowers M365 Copilot to access app metadata, Dataverse schemas, relationships, and permissions. Learn how MCP enables Copilot to answer natural-language queries, aggregate data, and deliver real-time insights, plus best practices for Copilot-ready apps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772578847951001DH4K","RegistrantKey":"1768245294609build26-1772578847951001DH4K"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"e63c4f3a-61a4-4eb0-918d-cd7f687814c7","startDateTime":"2026-06-03T17:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"John Link, Viktor Veis","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779478435395001BFO1","langLocale":"en-US","sessionId":"1779478435395001BFO1","sessionInstanceId":"1779478435395001BFO1","sessionCode":"DEM315","sessionCodeNormalized":"EN-US-DEM315","title":"Embracing frontier R&D with Microsoft Discovery","sortTitle":"embracing frontier r&d with microsoft discovery","description":"Microsoft Discovery is bringing Agentic Discovery to life.\u00A0 Much as agentic software engineering is not just vibe coding, Agentic Discovery is not just equipping researchers with AI tools.\u00A0 Agentic Discovery is a new operating model where autonomous teams of AI agents - guided by human expertise - execute entire cycles of the scientific method, from hypothesis formation to experimentation and reasoning.\u00A0 Learn how Microsoft Discovery is making this vision reality today.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with John Link introducing himself as a Partner Product Manager for Microsoft Discovery, welcoming the audience and outlining the agenda <a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>. He explains that the talk will focus first on why the Microsoft Discovery platform is significant, followed by a live demonstration by Victor. John sets the stage by describing Microsoft Discovery as a technology aimed at revolutionizing R&D across diverse fields such as chemistry, materials science, biotechnology, and physics, all powered by AI-driven approaches to accelerate scientific breakthroughs <a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>.</p>\n\n<p><strong>AI Inflection and Frontier R&D:</strong> John then elaborates on the broader AI landscape, pointing out that we are at a pivotal moment where AI is evolving from assisting humans to performing tasks autonomously <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:39\">00:01:39</a>. He notes that this shift doesn’t replace humans but redefines their role to guide and evaluate outcomes. Drawing parallels with software engineering’s transition through GitHub Copilot since 2021, he highlights how agentic models have fundamentally altered developer workflows and productivity. In November 2023, new models like GPT-4.5 and GPT-5.2 introduced advanced AI agency, crossing the threshold where systems could autonomously generate entire solutions—a transformation John predicts is now coming to scientific discovery <a class=\"medius-aisummary\" data-starttime=\"00:02:17\">00:02:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:48\">00:02:48</a>. He introduces “agentic discovery,” where teams of AI agents continuously run end-to-end scientific loops, guided by expert human intent, marking the beginning of what Microsoft calls “Frontier R&D” <a class=\"medius-aisummary\" data-starttime=\"00:03:28\">00:03:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:57\">00:03:57</a>.</p>\n\n<p><strong>Introducing Microsoft Discovery and the New App:</strong> John explains that Microsoft Discovery enables AI-driven scientific processes at scale, where thousands or even millions of hypotheses can be explored in parallel <a class=\"medius-aisummary\" data-starttime=\"00:04:23\">00:04:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:45\">00:04:45</a>. He recaps prior milestones, noting its introduction at Build as a private preview and its current general availability <a class=\"medius-aisummary\" data-starttime=\"00:04:50\">00:04:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:54\">00:04:54</a>. To engage more users, Microsoft now offers a complementary lightweight tool, the Microsoft Discovery app, which only requires a GitHub Copilot license and provides an easy entry point into Frontier R&D <a class=\"medius-aisummary\" data-starttime=\"00:05:23\">00:05:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a>. John transitions to Victor to demonstrate the app’s core features, including the Discovery Engine and the innovative scientific Bookshelf system.</p>\n\n<p><strong>Live Demonstration by Victor:</strong> Victor begins his demo by walking through the installation process using the GitHub repository and logging into the local Discovery app <a class=\"medius-aisummary\" data-starttime=\"00:06:07\">00:06:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:48\">00:06:48</a>. He highlights three main features: the Bookshelf (a graph-based knowledge retrieval system), the Discovery Engine (which automates hypothesis generation and testing), and a scientific interface built on Visual Studio Code <a class=\"medius-aisummary\" data-starttime=\"00:07:18\">00:07:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:36\">00:07:36</a>. Demonstrating a real project, Victor performs an analysis comparing serverless and container architectures for AI workloads. Using indexed research papers, the system produces structured Jupyter notebooks with cost-performance tradeoffs and hybrid execution recommendations <a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. The app also summarizes findings in executive one-pagers aimed at decision-makers. He then showcases the Discovery Engine’s extended capabilities for autonomous task execution, explaining how agents create, assign, and monitor tasks across iterative experimental loops <a class=\"medius-aisummary\" data-starttime=\"00:15:47\">00:15:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>. This closed-loop system can operate continuously for hours or days, refining insights automatically.</p>\n\n<p><strong>Real-World Use Cases and Customer Examples:</strong> Returning to the stage, John contextualizes the Discovery Engine with a visual of agents autonomously conducting experiments following scientific procedures <a class=\"medius-aisummary\" data-starttime=\"00:20:04\">00:20:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:33\">00:20:33</a>. He then shares a practical collaboration with a customer, “Science Co,” whom Microsoft assisted in determining optimal R&D investments. Using business-focused AI agents, Science Co identified heat transfer fluids for semiconductors as a high-potential area <a class=\"medius-aisummary\" data-starttime=\"00:21:05\">00:21:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:14\">00:21:14</a>. The Discovery platform was used to analyze scientific data, model molecular properties, and propose synthesis pathways, leading to four new candidate materials <a class=\"medius-aisummary\" data-starttime=\"00:22:18\">00:22:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:33\">00:22:33</a>. These outcomes demonstrate how Discovery accelerates research cycles while preserving reproducibility. John emphasizes the role of autonomous agentic systems in real-world innovation across materials, biology, and energy research.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> Closing the session, John highlights the broader adoption of Microsoft Discovery by organizations like BHP, Cambridge Consultants, and Silico Trials <a class=\"medius-aisummary\" data-starttime=\"00:23:19\">00:23:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:38\">00:23:38</a>. He directs viewers to Microsoft Discovery documentation and the GitHub repository for immediate access <a class=\"medius-aisummary\" data-starttime=\"00:23:44\">00:23:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:55\">00:23:55</a>, encouraging participants to experiment with the platform and provide feedback. The session concludes with gratitude to the audience and an invitation to be part of this evolving ecosystem that converges AI agency, scientific reasoning, and scalable computation to redefine the pace and potential of modern discovery <a class=\"medius-aisummary\" data-starttime=\"00:24:05\">00:24:05</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:00 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","speakerIds":["1768245294609build26-1778019408222001qSEO","1768245294609build26-1779486927569001tnkN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.236Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/fdbf653e-2bbd-4ef5-bcdc-d3ff526ebd98","ogTitle":"Embracing frontier R&D with Microsoft Discovery","ogDescription":"Microsoft Discovery is bringing Agentic Discovery to life.\u00A0 Much as agentic software engineering is not just vibe coding, Agentic Discovery is not just equipping researchers with AI tools.\u00A0 Agentic Discovery is a new operating model where autonomous teams of AI agents - guided by human expertise - execute entire cycles of the scientific method, from hypothesis formation to experimentation and reasoning.\u00A0 Learn how Microsoft Discovery is making this vision reality today.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778019408222001qSEO","RegistrantKey":"1768245294609build26-1778019408222001qSEO"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779486927569001tnkN","RegistrantKey":"1768245294609build26-1779486927569001tnkN"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"}]},{"@search.score":1.0,"documentId":"4370d49a-a158-4421-981e-ca9fb135b743","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nikola Metulev, Lei Xu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018718946001r6o5","langLocale":"en-US","sessionId":"1774018718946001r6o5","sessionInstanceId":"1774018718946001r6o5","sessionCode":"DEM345","sessionCodeNormalized":"EN-US-DEM345","title":"From prompt to app, build AI powered apps on Windows","sortTitle":"from prompt to app, build ai powered apps on windows","description":"From prompt to app, see how you can use GitHub Copilot CLI and skills to build an AI powered application that takes advantage of local compute on your Windows dev box. \nSee an E2E workflow for building a high‑performance desktop application that combines native UI and HW‑accelerated local compute for running AI workloads with Windows ML and Windows AI APIs.  See how you can use the Windows ML model kit to prepare the models to run with the highest performance on your local Windows device.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The presenters begin by greeting the audience at <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a> and introducing the session topic — building Windows applications using Copilot and agents, and integrating AI into those apps. They explain that instead of a typical slide presentation, the session will feature live demos primarily run from the terminal. One speaker focuses on Windows AI platforms, while the other specializes in skills and agents, creating a complementary perspective for the discussion. They promise interactive demonstrations, inviting attendees to follow along and ask questions about developing smarter Windows applications powered by AI.</p>\n\n<p><strong>Demo: Building a Sample AI Application with Copilot:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>, the speakers showcase a sample Windows application created entirely using GitHub Copilot and WinUI. This app processes media files offline, performing AI-driven tasks such as generating transcripts (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>), creating summaries, and embedding captions into videos. The local AI stack, leveraging Whisper and Microsoft Foundry, allows offline transcription and on-device inference. They highlight how Foundry and ONNX runtime help deploy AI models, even large ones like a 14-billion parameter transformer running locally on GPUs. Impressively, the app was fully built by Copilot within about one day (<a class=\"medius-aisummary\" data-starttime=\"00:02:45\">00:02:45</a>), underscoring how Copilot accelerates development by eliminating manual coding for much of the process.</p>\n\n<p><strong>Exploring Copilot Plugins and WinUI Skills:</strong> The session transitions at <a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a> to demonstrating Copilot’s plugin system. The WinUI plugin includes skills for design, packaging, code review, and setup (<a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>), enabling agents to autonomously build apps. They point out that the skills are simple markdown files but come bundled with CLI tools empowering agents to design and test real applications. Additionally, new WinUI templates are introduced (<a class=\"medius-aisummary\" data-starttime=\"00:04:43\">00:04:43</a>), which simplify manifest and XAML file creation, making CLI-based builds possible without Visual Studio. The presenters also highlight the new “Winapp” CLI tool (<a class=\"medius-aisummary\" data-starttime=\"00:06:10\">00:06:10</a>) that allows running and packaging applications directly from the command line — a capability previously unavailable. These tools enable agents and humans alike to explore APIs, identify compatible controls, and search component samples quickly via a unified workflow.</p>\n\n<p><strong>Optimizing AI Models with WinML:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:09:38\">00:09:38</a>, the presenters unveil a newly released tool — WinML CLI. This utility streamlines converting, optimizing, and deploying machine learning models for Windows. Using a Hugging Face vision transformer model as an example, they walk through exporting it (<a class=\"medius-aisummary\" data-starttime=\"00:13:08\">00:13:08</a>), analyzing its operators to identify hardware optimization opportunities (<a class=\"medius-aisummary\" data-starttime=\"00:13:56\">00:13:56</a>), and applying automatic performance improvements. These steps, executed with commands like `winml analyze` and `winml optimize`, fuse operators and reduce compute complexity. They demonstrate how the final optimized model performs rapid inference on the NPU, achieving roughly 3-millisecond latency (<a class=\"medius-aisummary\" data-starttime=\"00:18:19\">00:18:19</a>) and supporting over 300 frames per second — an impressive feat for local processing on Copilot Plus devices.</p>\n\n<p><strong>Live Use Case: Searching Video Frames with AI:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:18:35\">00:18:35</a>, the presenters demonstrate how the optimized model powers a new “Find Frames” feature. Built in about 25 minutes through Copilot, the feature scans long videos — in this example, a two-hour keynote — for specific items or people using text prompts. By generating embeddings across frames and comparing them with a search term (“leather jacket”), the application locates the desired frames efficiently, showing a match with NVIDIA’s CEO’s appearance. The entire operation runs offline on the NPU, showcasing the power of AI-accelerated workloads integrated with Copilot-driven app development.</p>\n\n<p><strong>Recap and Closing Remarks:</strong> In the terminal-based presentation recap at <a class=\"medius-aisummary\" data-starttime=\"00:20:16\">00:20:16</a>, they summarize how Windows agents and developers both benefit from advanced tooling — matching skills, templates, and AI utilities. Agents can now build, test, and optimize applications from the command line just like humans, accelerating the creation of Windows apps that leverage AI. The discussion closes with reflections on using Copilot to create presentations and applications quickly, showing that app development is becoming as simple as asking a capable AI assistant for help. Finally, at <a class=\"medius-aisummary\" data-starttime=\"00:22:10\">00:22:10</a>, they thank the audience and encourage further exploration through provided links, marking the end of an informative and hands-on session.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/851eb058-a594-4193-8437-6010e4400b02","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/851eb058-a594-4193-8437-6010e4400b02","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/851eb058-a594-4193-8437-6010e4400b02","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/851eb058-a594-4193-8437-6010e4400b02","speakerIds":["1768245294609build26-1774886013101001R9mk","1768245294609build26-1774385980999001nI1l"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.858Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/851eb058-a594-4193-8437-6010e4400b02","ogTitle":"From prompt to app, build AI powered apps on Windows","ogDescription":"From prompt to app, see how you can use GitHub Copilot CLI and skills to build an AI powered application that takes advantage of local compute on your Windows dev box. \nSee an E2E workflow for building a high‑performance desktop application that combines native UI and HW‑accelerated local compute for running AI workloads with Windows ML and Windows AI APIs.  See how you can use the Windows ML model kit to prepare the models to run with the highest performance on your local Windows device.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"C++","logicalValue":"C++"},{"displayValue":"C","logicalValue":"C"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774886013101001R9mk","RegistrantKey":"1768245294609build26-1774886013101001R9mk"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774385980999001nI1l","RegistrantKey":"1768245294609build26-1774385980999001nI1l"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM345","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"3b37ca02-aaaa-4c74-b22a-29050e3fcbe6","startDateTime":"2026-06-03T20:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Freek Berson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776289484544001DLd6","langLocale":"en-US","sessionId":"1776289484544001DLd6","sessionInstanceId":"1776289484544001DLd6","sessionCode":"DEM363","sessionCodeNormalized":"EN-US-DEM363","title":"From the Field: Accelerate Your Development for Microsoft Marketplace","sortTitle":"from the field: accelerate your development for microsoft marketplace","description":"Gain practical insights and proven strategies for your development journey for Microsoft Marketplace. Streamline your solutions' development, packaging, and deployment while ensuring compliance with Microsoft’s requirements. Discover common pitfalls, best practices, and automation techniques to reduce time-to-market. Whether you're new to developing for the Microsoft Marketplace or looking to optimize your approach, this session delivers real-world guidance to help you develop and publish efficiently.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The video opens with a warm greeting from Craig Berson, a Microsoft MVP based in the Netherlands, who sets an upbeat tone for the session on creating and publishing Azure infrastructure solutions. At <a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>, Craig emphasizes that while building Azure solutions with services like GitHub Copilot and automation tools is easy, delivering trusted, scalable deployments through customer-facing channels remains challenging. His goal, introduced around <a class=\"medius-aisummary\" data-starttime=\"00:00:42\">00:00:42</a>, is to guide viewers through a journey of transforming infrastructure-as-code templates into deployable products in the Microsoft Marketplace, starting from template creation to packaging and publishing.</p>\n\n<p><strong>Azure Marketplace Benefits and Offer Types:</strong> Craig transitions at <a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a> into explaining the benefits of the Marketplace for independent software vendors (ISVs), partners, and consumers. He highlights its role in improving visibility, simplifying reselling, and aligning consumption commitments for Azure customers. By <a class=\"medius-aisummary\" data-starttime=\"00:02:36\">00:02:36</a>, he introduces the session’s focal point—the Azure Application offer—which lets creators package reusable infrastructure templates with different tiers (silver, gold, platinum) for customer deployment flexibility. Craig explains how offers deploy managed applications and resource groups (<a class=\"medius-aisummary\" data-starttime=\"00:03:26\">00:03:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:47\">00:03:47</a>), connecting telemetry and resources through managed groups for organized and scalable delivery to end users.</p>\n\n<p><strong>Technical Foundation and Template Workflow:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>, Craig demystifies the technical mechanics of Marketplace offers, describing them essentially as ZIP files containing mandatory components: the ARM template and the Create UI definition. He encourages developers to author their infrastructure using Bicep (<a class=\"medius-aisummary\" data-starttime=\"00:05:03\">00:05:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>), taking advantage of Visual Studio Code extensions and the Bicep visualizer for ease of development and visualization. Because the Marketplace only supports ARM templates natively, Craig demonstrates transpiling Bicep into ARM using the \"az bicep build\" command at <a class=\"medius-aisummary\" data-starttime=\"00:06:43\">00:06:43</a>, advocating for testing deployments thoroughly before integration (<a class=\"medius-aisummary\" data-starttime=\"00:07:58\">00:07:58</a>). His advice: start with concise, readable Bicep syntax, then convert for Marketplace compatibility to maintain clean code and efficiency.</p>\n\n<p><strong>UI Design and Template Mapping:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:08:25\">00:08:25</a>, Craig dives into building and testing the user interface experience through the Create UI definition file. He introduces the concept of the sandbox environment (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:37\">00:09:37</a>), which allows developers to design and preview form elements such as tabs, dropdowns, and configuration parameters used during deployment in the Azure portal. Craig demonstrates mapping UI parameters precisely to Bicep/ARM template parameters (<a class=\"medius-aisummary\" data-starttime=\"00:10:33\">00:10:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:47\">00:11:47</a>), stressing exact one-to-one, case-sensitive correspondence. This step ensures the Marketplace deployment interface accurately triggers infrastructure creation defined in code. He concludes this part with packaging guidance—adding both template and UI files into the ZIP—and a pre-upload validation using the ARM TTK Toolkit (<a class=\"medius-aisummary\" data-starttime=\"00:12:07\">00:12:07</a>), which runs 49 tests to verify compliance, security, and syntax adherence before submission.</p>\n\n<p><strong>Partner Center Configuration and Deployment Result:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>, Craig walks through publishing via Partner Center, showing how to upload the ZIP, version it correctly, and select deployment modes (incremental vs. complete). He discusses notification URL settings (<a class=\"medius-aisummary\" data-starttime=\"00:15:28\">00:15:28</a>) for monitoring purchases, cancellations, and upgrades and configuring resource group permissions to control publisher and customer access (<a class=\"medius-aisummary\" data-starttime=\"00:16:02\">00:16:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:34\">00:16:34</a>). Craig then demonstrates actual Marketplace deployment results (<a class=\"medius-aisummary\" data-starttime=\"00:16:48\">00:16:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:27\">00:17:27</a>), inspecting managed applications and resource groups populated with Azure assets like VMs, load balancers, and key vaults. He explains nested deployments (<a class=\"medius-aisummary\" data-starttime=\"00:18:04\">00:18:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:16\">00:18:16</a>) that grant ISVs credit for customer-side deployments, rounding out the real-world Marketplace implementation process.</p>\n\n<p><strong>Summary, Best Practices, and Closing Remarks:</strong> In the final segment (<a class=\"medius-aisummary\" data-starttime=\"00:18:28\">00:18:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:41\">00:19:41</a>), Craig reviews the publishing workflow, reiterating the importance of local testing to cut down on iterative upload failures. His call-to-action encourages attendees to explore the Microsoft Marketplace booth for additional resources and expert discussions (<a class=\"medius-aisummary\" data-starttime=\"00:19:07\">00:19:07</a>). He concludes by sharing GitHub access to all sample code, promoting community engagement on Azure development, and offering giveaways of his Bicep book and event swag to inspire further learning. The talk ends on a motivational note, connecting Marketplace innovation with the spirit of collaboration and streamlined DevOps delivery across the Azure ecosystem.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"20:30 - 20:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/53a72f4d-0afb-4321-b633-4e8de5b474d3","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/53a72f4d-0afb-4321-b633-4e8de5b474d3","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/53a72f4d-0afb-4321-b633-4e8de5b474d3","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/53a72f4d-0afb-4321-b633-4e8de5b474d3","speakerIds":["1768245294609build26-1776289702978001L6Ff"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.902Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/53a72f4d-0afb-4321-b633-4e8de5b474d3","ogTitle":"From the Field: Accelerate Your Development for Microsoft Marketplace","ogDescription":"Gain practical insights and proven strategies for your development journey for Microsoft Marketplace. Streamline your solutions' development, packaging, and deployment while ensuring compliance with Microsoft’s requirements. Discover common pitfalls, best practices, and automation techniques to reduce time-to-market. Whether you're new to developing for the Microsoft Marketplace or looking to optimize your approach, this session delivers real-world guidance to help you develop and publish efficiently.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776289702978001L6Ff","RegistrantKey":"1768245294609build26-1776289702978001L6Ff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM363","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"1cc7679e-fa1a-4b9d-84b3-62b67e01d331","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aamir Jawaid, Umang Sehgal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018702063001rqph","langLocale":"en-US","sessionId":"1774018702063001rqph","sessionInstanceId":"1774018702063001rqph","sessionCode":"DEM332","sessionCodeNormalized":"EN-US-DEM332","title":"From zero to teammate in 25 minutes: Build a Teams agent live","sortTitle":"from zero to teammate in 25 minutes: build a teams agent live","description":"In 25 minutes, flat, watch an agent come to life inside Microsoft Teams, one that works alongside you in the channels and chats you already use. Using the Teams CLI, the Teams SDK's new skills plugin, and GitHub Copilot, we'll go from blank terminal to a deployed agent your colleagues can @mention, delegate to, and collaborate with. No slideware, no shortcuts.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The session opens with the host, Umang, welcoming attendees to the Build conference (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>). He introduces himself as a senior product manager for Microsoft Teams and his co-presenter, Aamir, a senior software engineer on the Teams SDK. The agenda includes a quick recap of the Teams SDK’s previous developments, the introduction of a new command-line interface (CLI), a live demonstration of building an agent in under 25 minutes, and the process of transforming that agent into a fully integrated teammate with its own identity inside Teams (<a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>).</p>\n\n<p><strong>SDK Evolution and Simplification Goals:</strong> Umang explains that last year’s Build introduced the Teams AI library in TypeScript and C#, while this year’s focus is on the new integrated Teams SDK (<a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>). This SDK consolidates tools and libraries into a unified developer experience and extends functionality to group collaboration. He reviews partnerships with various companies that have integrated with Teams and underscores the goal of simplifying agent development. Umang describes three core pillars of simplification: enabling agents built in any development stack to run in Teams, minimizing setup friction through automated scaffolding, and allowing agents to become true \"teammates\" with individual identities (<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>).</p>\n\n<p><strong>Live Demo – Building and Integrating the Agent:</strong> Aamir begins a live demonstration, showing a “project management helper” web application that summarizes and tracks project statuses (<a class=\"medius-aisummary\" data-starttime=\"00:04:03\">00:04:03</a>). He uses the new Teams CLI to register and configure the agent with the command \"Teams app create,\" explaining that the CLI automatically creates relevant app and bot resources and provides installation links (<a class=\"medius-aisummary\" data-starttime=\"00:05:25\">00:05:25</a>). He then shows how Copilot can streamline integration using natural language prompts like “Integrate this application into Microsoft Teams.” The CLI is demonstrated in both interactive and JSON modes, offering ease-of-use for both human developers and agents. Features like progressive command disclosure help developers and AI agents efficiently find only the relevant commands they need (<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>).</p>\n\n<p><strong>Deploying and Demonstrating the Agent in Teams:</strong> After configuring the agent, Aamir installs it within Teams using the provided link, showcasing that the same query used on the web app — asking for a project summary — now works seamlessly inside Teams (<a class=\"medius-aisummary\" data-starttime=\"00:12:11\">00:12:11</a>). The agent also functions in a group chat, responding to project-related queries across participants. Umang rejoins to emphasize that the CLI is “agent-first,” allowing seamless automation for coding and deployment tasks. He previews upcoming group UX features such as emoji reactions, targeted messages, slash commands, quoted replies, and markdown support, enhancing collaborative experiences with intelligent agents inside Teams (<a class=\"medius-aisummary\" data-starttime=\"00:14:10\">00:14:10</a>).</p>\n\n<p><strong>Transforming Agents into True Teammates:</strong> The presenters shift focus to building agents that act as full-fledged teammates across Microsoft 365. Umang explains the concept of “Agents 365” — giving each agent an identity within the M365 ecosystem so they can be mentioned in chats, tagged in documents, or even send emails just like human coworkers (<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>). A video demo shows how the same project management agent now functions as a “mobile redesign project manager” capable of one-on-one interactions, contextual awareness of its assigned project, and inclusion in relevant group discussions. Thanks to its identity, the agent can send messages and emails directly from its own address inside Outlook, linking collaboration across Teams and email environments (<a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>).</p>\n\n<p><strong>Conclusion and Call to Action:</strong> The session concludes by reinforcing that the new Teams SDK and CLI dramatically simplify building intelligent, collaborative agents that feel like real teammates within Microsoft’s productivity suite (<a class=\"medius-aisummary\" data-starttime=\"00:19:53\">00:19:53</a>). Developers are encouraged to explore the SDK and CLI via the official Microsoft link, attend the follow-up session on group chat integrations, and leverage the evolving tools to create agent-driven experiences within Teams. Umang thanks attendees, closing on the theme that Teams is becoming the central workspace where both humans and intelligent agents can work side by side efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:20:02\">00:20:02</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:10 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:35:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/901bde6f-0064-4018-891f-4b8fbaba8d71","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/901bde6f-0064-4018-891f-4b8fbaba8d71","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/901bde6f-0064-4018-891f-4b8fbaba8d71","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/901bde6f-0064-4018-891f-4b8fbaba8d71","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/901bde6f-0064-4018-891f-4b8fbaba8d71","speakerIds":["1768245294609build26-1778598527819001cKDs","1768245294609build26-1775594818790001AuRe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.457Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/901bde6f-0064-4018-891f-4b8fbaba8d71","ogTitle":"From zero to teammate in 25 minutes: Build a Teams agent live","ogDescription":"In 25 minutes, flat, watch an agent come to life inside Microsoft Teams, one that works alongside you in the channels and chats you already use. Using the Teams CLI, the Teams SDK's new skills plugin, and GitHub Copilot, we'll go from blank terminal to a deployed agent your colleagues can @mention, delegate to, and collaborate with. No slideware, no shortcuts.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778598527819001cKDs","RegistrantKey":"1768245294609build26-1778598527819001cKDs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775594818790001AuRe","RegistrantKey":"1768245294609build26-1775594818790001AuRe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM332","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"7477546d-8f79-4aca-8ab8-59b658b11951","startDateTime":"2026-06-02T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ari LiVigni, Alejandro Menocal","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774041861148001j83D","langLocale":"en-US","sessionId":"1774041861148001j83D","sessionInstanceId":"1774041861148001j83D","sessionCode":"DEM350","sessionCodeNormalized":"EN-US-DEM350","title":"GitHub Agentic Workflows: Automation That Actually Reads the Room","sortTitle":"github agentic workflows: automation that actually reads the room","description":"GitHub Agentic Workflows let your repo improve itself. With a simple markdown file and one command, GitHub Actions launches an AI agent to triage issues, fix CI failures, update docs, and improve tests, with no complex YAML required. See a live demo from minimal workflow file to a safe, sandboxed pipeline that delivers a ready‑to‑review PR. Your repo on autopilot, with you in control.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session opens with greetings from Ari Livigni and Alejandro Menocal, both senior professionals from GitHub, welcoming attendees to Microsoft Build Day One <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>. They introduce their discussion topic—GitHub Agentic Workflows—highlighting how automation powered by AI can dynamically \"read the room\" and adapt processes intelligently <a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>. The presenters emphasize how these workflows simplify traditional developer tasks by allowing workflow definitions in Markdown rather than YAML, with GitHub Copilot assisting in generation and compilation <a class=\"medius-aisummary\" data-starttime=\"00:00:33\">00:00:33</a>. This shift enables developers to automate continuous integration and delivery (CI/CD) easily, reducing syntax complexity while increasing flexibility in routine project management.</p>\n\n<p><strong>Capabilities and Ecosystem:</strong> The speakers discuss the benefits of using Agentic Workflows for development automation, such as triaging test failures, generating periodic reports or issues, and streamlining pull requests <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>. They explain how various AI models—Copilot, Claude, CodeX, or Gemini—can dynamically diagnose issues, propose fixes, or update dependencies <a class=\"medius-aisummary\" data-starttime=\"00:01:34\">00:01:34</a>. Alejandro highlights the ecosystem’s flexibility and mentions the upcoming public preview release planned for the following week, inviting participants to experiment through GitHub’s website and use a new skills exercise to explore workflow capabilities firsthand <a class=\"medius-aisummary\" data-starttime=\"00:01:58\">00:01:58</a>. This exercise demonstrates how everyday coding tasks can become more efficient thanks to generative automation and multi-model integration.</p>\n\n<p><strong>Hands-On Skills Exercise Introduction:</strong> The presenters lead into a GitHub Skills exercise, using a demo project centered around a fictitious “Mona” website that aggregates GitHub blog posts, changelogs, and notes into regular updates <a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>. The Agentic Workflow in this context is designed to automatically generate pull requests for content refreshes, which can then be peer-reviewed and merged <a class=\"medius-aisummary\" data-starttime=\"00:02:35\">00:02:35</a>. The exercise introduces participants to repository setup steps, including using GitHub Skills to install the workflows and scaffolding the environment. As seen in the demo, when the agent installs, it produces supporting configuration files—such as the `.agent` definition and the custom “skills” files—under the `.github` directory, illustrating modular and transparent automation <a class=\"medius-aisummary\" data-starttime=\"00:03:25\">00:03:25</a>.</p>\n\n<p><strong>Demonstration of Workflow Creation:</strong> The live demo continues with Ari and Alejandro walking through command-line steps to initialize the repository, run commands that create pull requests, merge setup files, and automatically scaffold related code <a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>. They explain that this process mimics the GitHub Copilot setup where foundational structures are autogenerated, leaving room for developers to customize and expand based on their CI/CD or automation goals <a class=\"medius-aisummary\" data-starttime=\"00:05:07\">00:05:07</a>. As the workflow runs, the presenters demonstrate creating a new branch and instructing the AI agent to build automation that keeps the demo website updated with blog and changelog data. The workflow, authored in Markdown, is compiled into YAML “lock” files, which makes them executable as standardized GitHub Actions <a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>.</p>\n\n<p><strong>Customization, Security, and Compilation:</strong> Alejandro elaborates on how agents can be modified for unique needs—such as scheduling updates or customizing triggers within the repository <a class=\"medius-aisummary\" data-starttime=\"00:07:29\">00:07:29</a>. A key security concept presented is the use of “safe outputs,” ensuring that the agent itself only has read access while actions requiring write permissions (like opening pull requests) are delegated to secure workflow steps <a class=\"medius-aisummary\" data-starttime=\"00:08:39\">00:08:39</a>. This prevents unauthorized changes, protecting codebases from AI hallucinations or deletions. The presenters then show how the workflow generates front matter and markdown instructions that are compiled into executable logic, forming a secure and human-readable automation layer <a class=\"medius-aisummary\" data-starttime=\"00:13:29\">00:13:29</a>.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> The session wraps up with the agents successfully producing an updatable GitHub Info file (<a class=\"medius-aisummary\" data-starttime=\"00:11:45\">00:11:45</a>) and compilation of its YAML lock file (<a class=\"medius-aisummary\" data-starttime=\"00:16:08\">00:16:08</a>). This lock file acts as the actual GitHub Actions workflow that executes regularly or on-demand. Ari emphasizes that Agentic Workflows extend GitHub Actions’ capabilities to a new paradigm, allowing developers to control automation in natural language <a class=\"medius-aisummary\" data-starttime=\"00:12:13\">00:12:13</a>. The presenters close by inviting attendees to scan a QR code and try the GitHub Skills exercise themselves <a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>, promising continuous iteration and improvements to the experience as Agentic Workflows evolve.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:40 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/ce603017-0583-4aa8-85a6-a7fb79aaa496","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ce603017-0583-4aa8-85a6-a7fb79aaa496","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ce603017-0583-4aa8-85a6-a7fb79aaa496","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ce603017-0583-4aa8-85a6-a7fb79aaa496","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ce603017-0583-4aa8-85a6-a7fb79aaa496","speakerIds":["1768245294609build26-1767637462082001kD5b","1768245294609build26-1769537807790001qHZB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.859Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ce603017-0583-4aa8-85a6-a7fb79aaa496","ogTitle":"GitHub Agentic Workflows: Automation That Actually Reads the Room","ogDescription":"GitHub Agentic Workflows let your repo improve itself. With a simple markdown file and one command, GitHub Actions launches an AI agent to triage issues, fix CI failures, update docs, and improve tests, with no complex YAML required. See a live demo from minimal workflow file to a safe, sandboxed pipeline that delivers a ready‑to‑review PR. Your repo on autopilot, with you in control.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767637462082001kD5b","RegistrantKey":"1768245294609build26-1767637462082001kD5b"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769537807790001qHZB","RegistrantKey":"1768245294609build26-1769537807790001qHZB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM350","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"bfc77474-54d5-407e-8659-f12a46914c07","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ellie Bennett, Denizhan Yigitbas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710574001rWWg","langLocale":"en-US","sessionId":"1774018710574001rWWg","sessionInstanceId":"1774018710574001rWWg","sessionCode":"DEM305","sessionCodeNormalized":"EN-US-DEM305","title":"GitHub Copilot Anywhere: From Remote Control CLIs to Cloud Sandboxes","sortTitle":"github copilot anywhere: from remote control clis to cloud sandboxes","description":"Modern development no longer happens in one place. You might start in your terminal, step away, and pick things back up from another device or let work continue in the cloud. This session explores using GitHub Copilot across environments, from local CLI sessions to full cloud execution with Remote Sandboxes. See how Copilot runs commands, iterates on changes, and continues work without losing context, while you stay in control.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Setup:</strong> The session opens with welcoming remarks and a brief introduction to the talk titled \"GitHub Copilot Anywhere\" presented by Ellie and Dennison (<a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>). After a few microphone checks (<a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>), Ellie sets the tone by emphasizing the fast pace of innovation and how developer creativity often strikes outside of traditional desk environments, like at airports or in living rooms (<a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:48\">00:01:48</a>). This idea naturally introduces the theme of developer flexibility—being able to code anywhere and anytime using GitHub Copilot.</p>\n\n<p><strong>Remote Control Overview and Demo:</strong> Ellie introduces GitHub Copilot’s new remote control feature, describing how it enables developers to take their local work on the go and interact with it remotely through devices or github.com (<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:56\">00:02:56</a>). A live demo follows featuring a custom World Cup app example where Ellie uses Copilot to improve the app’s UI while detecting bugs like missing host countries (<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>). She activates \"remote on,\" generates a live link and QR code, scans it with her phone, and demonstrates real-time synchronization between her CLI and mobile device (<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>). Permissions appear on the phone, confirming the remote capabilities. Ellie closes this section with usage tips: enabling remote sessions by default, keeping terminals online, and coordinating enterprise policies (<a class=\"medius-aisummary\" data-starttime=\"00:08:38\">00:08:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:36\">00:09:36</a>).</p>\n\n<p><strong>Introducing Cloud Sandboxes:</strong> Transitioning from remote control, Dennison expands on the concept of Copilot having its own environment independent of the local machine. He unveils \"Cloud Sandbox\" for Copilot, now in public preview (<a class=\"medius-aisummary\" data-starttime=\"00:10:22\">00:10:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:45\">00:10:45</a>). In this feature, each Copilot session operates inside a secure, isolated cloud environment, allowing uninterrupted access even if the local machine shuts down (<a class=\"medius-aisummary\" data-starttime=\"00:11:13\">00:11:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:27\">00:11:27</a>). Developers can allocate cloud resources to handle heavy workloads without straining local systems. He demonstrates switching from a regular CLI session to a sandboxed cloud session using a simple \"dash dash cloud\" command (<a class=\"medius-aisummary\" data-starttime=\"00:14:06\">00:14:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:38\">00:14:38</a>), confirming that once launched, Copilot instances persist online regardless of host activity.</p>\n\n<p><strong>Cloud Sandbox Demo and Policy Controls:</strong> In a hands-on example, Dennison builds upon the World Cup app by using a sandbox to add a \"live broadcasting\" feature—showing minute-by-minute game updates for the final (<a class=\"medius-aisummary\" data-starttime=\"00:12:34\">00:12:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:12\">00:13:12</a>). After launching the session in the cloud, he closes his tab to demonstrate persistence; the Copilot interaction continues because it's hosted remotely (<a class=\"medius-aisummary\" data-starttime=\"00:15:31\">00:15:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:36\">00:15:36</a>). Policies governing this environment—like firewall rules and repository scopes—can be managed under settings for Copilot’s cloud agents (<a class=\"medius-aisummary\" data-starttime=\"00:15:49\">00:15:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:05\">00:16:05</a>). Dennison animates the talk by simulating a World Cup final where Turkey triumphs, a playful test of the new broadcasting capabilities (<a class=\"medius-aisummary\" data-starttime=\"00:17:12\">00:17:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>).</p>\n\n<p><strong>Performance, Persistence, and Cost Efficiency:</strong> Dennison explains the inner workings of sandbox persistence and billing: whenever Copilot finishes its tasks, the sandbox’s state and disk snapshot are saved, reducing compute usage to zero until resumed (<a class=\"medius-aisummary\" data-starttime=\"00:18:19\">00:18:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:01\">00:19:01</a>). This design ensures cost efficiency and instant continuity upon reactivation. He summarizes the two working models—remote control sessions for local work requiring the host machine and cloud sandbox sessions granting Copilot full autonomy in isolated environments (<a class=\"medius-aisummary\" data-starttime=\"00:19:10\">00:19:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:41\">00:19:41</a>).</p>\n\n<p><strong>Chronicle Feature and Conclusion:</strong> Ellie closes the session by introducing \"Chronicle,\" a meta-feature that logs and indexes all user Copilot sessions for insights, tips, and enhancements across platforms like CLI, VS Code, JetBrains, and github.com (<a class=\"medius-aisummary\" data-starttime=\"00:19:57\">00:19:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:09\">00:21:09</a>). By running commands such as \"/chronicle tips,\" developers can access stand-up summaries, cost optimizations, and guidance based on prior usage (<a class=\"medius-aisummary\" data-starttime=\"00:21:49\">00:21:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:33\">00:22:33</a>). The presenters conclude by reaffirming Copilot’s evolving role in empowering developers to choose not only how they work but where and when work happens (<a class=\"medius-aisummary\" data-starttime=\"00:22:43\">00:22:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:53\">00:22:53</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:30 - 17:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c8128017-96b9-476c-9004-8ba91e2ef3cd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c8128017-96b9-476c-9004-8ba91e2ef3cd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c8128017-96b9-476c-9004-8ba91e2ef3cd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c8128017-96b9-476c-9004-8ba91e2ef3cd","speakerIds":["1768245294609build26-1774880387175001zXLI","1768245294609build26-1775278306002001tw5Z"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.79Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c8128017-96b9-476c-9004-8ba91e2ef3cd","ogTitle":"GitHub Copilot Anywhere: From Remote Control CLIs to Cloud Sandboxes","ogDescription":"Modern development no longer happens in one place. You might start in your terminal, step away, and pick things back up from another device or let work continue in the cloud. This session explores using GitHub Copilot across environments, from local CLI sessions to full cloud execution with Remote Sandboxes. See how Copilot runs commands, iterates on changes, and continues work without losing context, while you stay in control.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774880387175001zXLI","RegistrantKey":"1768245294609build26-1774880387175001zXLI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775278306002001tw5Z","RegistrantKey":"1768245294609build26-1775278306002001tw5Z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"ab4327da-78ab-479c-a25b-340a32f7f911","startDateTime":"2026-06-02T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nagkumar Arkalgud, Facundo Santiago","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018702220001rMvd","langLocale":"en-US","sessionId":"1774018702220001rMvd","sessionInstanceId":"1774018702220001rMvd","sessionCode":"DEM333","sessionCodeNormalized":"EN-US-DEM333","title":"How Foundry integrates with open-source frameworks and tools","sortTitle":"how foundry integrates with open-source frameworks and tools","description":"This demo walks through building a practical OpenClaw like agent using open-source technologies and then operationalizing the same solution in Microsoft Foundry. You will see how to connect enterprise tools with the Model Context Protocol, codify repeatable behavior with skills, add browser automation with Playwright CLI, observe with OpenTelemetry, and move from local development to cloud-hosted agents using open protocols like Responses API and A2A to agent-to-agent communication.","aiDescription":"<p><strong>Introduction and Objective:</strong> The video opens with Fakundo, Principal Product Manager at Microsoft, and Nakumar, Senior Software Engineer, introducing the session on how Microsoft Foundry integrates with open source frameworks and technologies (<a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>). They pose a key question: how can developers package and deploy agents built with open source tools to production without having to rewrite them (<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>)? They explain that the session will demonstrate building an agent — similar to well-known web-browsing assistants — using open source components and then deploying it on Microsoft Foundry to make it production-ready (<a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>).</p>\n\n<p><strong>Building a Basic Agent:</strong> Nakumar begins by describing the minimal setup needed to build a basic local agent: a model and an agent loop (<a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a>). The loop continuously queries the model for actions, executes tools when required, and determines next steps. Using LangChain’s chat model and a “create_deep_agent” function, the code remains compact and framework-consistent (<a class=\"medius-aisummary\" data-starttime=\"00:01:08\">00:01:08</a>). Since Foundry models expose OpenAI-compatible APIs, developers can point LangChain agents to Foundry without code rewrites, just by updating configuration. A demo shows this basic agent responding locally, confirming that it can converse but lacks functionality beyond model knowledge (<a class=\"medius-aisummary\" data-starttime=\"00:01:57\">00:01:57</a>).</p>\n\n<p><strong>Adding Tools with MCP Integration:</strong> To enable useful actions, the open source protocol MCP (Model Context Protocol) is introduced (<a class=\"medius-aisummary\" data-starttime=\"00:02:28\">00:02:28</a>). MCP allows an agent to discover and call external tools dynamically. In this demo, a Microsoft 365 integration called WorkIQ MCP server provides access to email tools such as search, get message details, and draft replies without the agent needing direct API knowledge (<a class=\"medius-aisummary\" data-starttime=\"00:02:49\">00:02:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>). The presenters run an MCP-enabled agent that connects to Microsoft 365 mail, retrieves emails, and executes these tools in real-time. This demonstrates the ability for LangGraph agents to access services like Outlook, Teams, or Calendar through an open protocol rather than proprietary SDKs (<a class=\"medius-aisummary\" data-starttime=\"00:03:33\">00:03:33</a>).</p>\n\n<p><strong>Introducing Skills and Web Interaction:</strong> The conversation shifts to enhancing agent intelligence using “skills” — reusable playbooks defining sequences of actions like inbox triage (<a class=\"medius-aisummary\" data-starttime=\"00:04:38\">00:04:38</a>). Nakumar explains that tools act as verbs, while skills embody workflows such as “triage inbox” or “draft replies.” Skills are simple markdown documents the agent reads when needed. A demo shows how skills improve agent behavior, transforming it from merely reading emails to classifying and prioritizing them based on content (<a class=\"medius-aisummary\" data-starttime=\"00:06:23\">00:06:23</a>). They also introduce web browsing capabilities powered by Playwright, an open source CLI tool that agents can use efficiently to fetch web data (e.g., checking product prices on Amazon) without the token overhead of large tool schemas (<a class=\"medius-aisummary\" data-starttime=\"00:07:07\">00:07:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:12\">00:08:12</a>).</p>\n\n<p><strong>Deploying and Monitoring on Foundry:</strong> The presenters move from local demos to production deployment using Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:09:23\">00:09:23</a>). Foundry hosts the same LangGraph agent as a managed service while exposing it via an OpenAI-compatible responses API. The deployment process includes setting up telemetry instrumentation before graph construction for observability. Foundry’s integration with Application Insights and OpenTelemetry allows inspection of model spans, latencies, and token usage to identify performance bottlenecks (<a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>). Since OpenTelemetry is an open standard, teams can visualize traces through any tool like Grafana or Azure Monitor. Demonstrations reveal trace replays showing user input, tool actions, and token costs across the workflow.</p>\n\n<p><strong>Agent-to-Agent Interoperability and Conclusion:</strong> Finally, they demonstrate cross-agent communication via the Agent-to-Agent (A2A) protocol (<a class=\"medius-aisummary\" data-starttime=\"00:13:09\">00:13:09</a>). Hosting the agent in Foundry automatically exposes an A2A endpoint, allowing other agents — such as one in Copilot CLI — to locate and invoke it dynamically through directory-based discovery and MCP-compatible tools like “search agent” and “call agent” (<a class=\"medius-aisummary\" data-starttime=\"00:14:09\">00:14:09</a>). The session concludes by emphasizing the power of open source interoperability: individual components (local agents, Foundry hosts, MCP servers) communicate seamlessly using shared standards without hard dependencies. Fakundo wraps up by inviting viewers to explore the accompanying repository and experiment with each stage of the showcased agent pipeline (<a class=\"medius-aisummary\" data-starttime=\"00:16:04\">00:16:04</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:00 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/60554504-db91-4d72-a3de-4f1c346b7ceb","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/60554504-db91-4d72-a3de-4f1c346b7ceb","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/60554504-db91-4d72-a3de-4f1c346b7ceb","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/60554504-db91-4d72-a3de-4f1c346b7ceb","speakerIds":["1768245294609build26-1776181207688001qvOn","1768245294609build26-1766703128481001X0K3"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.457Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/60554504-db91-4d72-a3de-4f1c346b7ceb","ogTitle":"How Foundry integrates with open-source frameworks and tools","ogDescription":"This demo walks through building a practical OpenClaw like agent using open-source technologies and then operationalizing the same solution in Microsoft Foundry. You will see how to connect enterprise tools with the Model Context Protocol, codify repeatable behavior with skills, add browser automation with Playwright CLI, observe with OpenTelemetry, and move from local development to cloud-hosted agents using open protocols like Responses API and A2A to agent-to-agent communication.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1766703128481001X0K3","RegistrantKey":"1768245294609build26-1766703128481001X0K3"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181207688001qvOn","RegistrantKey":"1768245294609build26-1776181207688001qvOn"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/DEM333","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"4be981cd-d1e3-4832-91aa-c77e5263103c","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeff Boudier, Osi Otugo, Vaidya Sambasivam","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018713721001rm1f","langLocale":"en-US","sessionId":"1774018713721001rm1f","sessionInstanceId":"1774018713721001rm1f","sessionCode":"DEM320","sessionCodeNormalized":"EN-US-DEM320","title":"Hugging Face open‑source models to production on Microsoft Foundry","sortTitle":"hugging face open‑source models to production on microsoft foundry","description":"Open-source models fuel modern AI but running them in production is hard. In this lightning talk, Microsoft and Hugging Face demonstrate how to deploy and scale Hugging Face models using Foundry Managed Compute inside Azure AI Foundry. Through a fast, end to end demo, developers will see how to go from model discovery to production inference without managing GPUs, while gaining autoscaling, governance, and enterprise grade performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:30 - 17:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:55:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/89dcba95-e116-4091-b82f-49452d243797","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/89dcba95-e116-4091-b82f-49452d243797","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/89dcba95-e116-4091-b82f-49452d243797","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/89dcba95-e116-4091-b82f-49452d243797","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/89dcba95-e116-4091-b82f-49452d243797","speakerIds":["1768245294609build26-1777488820140001xMpM","1768245294609build26-1777484162734001FjWO","1768245294609build26-1776978879433001zsDh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.853Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/89dcba95-e116-4091-b82f-49452d243797","ogTitle":"Hugging Face open‑source models to production on Microsoft Foundry","ogDescription":"Open-source models fuel modern AI but running them in production is hard. In this lightning talk, Microsoft and Hugging Face demonstrate how to deploy and scale Hugging Face models using Foundry Managed Compute inside Azure AI Foundry. Through a fast, end to end demo, developers will see how to go from model discovery to production inference without managing GPUs, while gaining autoscaling, governance, and enterprise grade performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776978879433001zsDh","RegistrantKey":"1768245294609build26-1776978879433001zsDh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777484162734001FjWO","RegistrantKey":"1768245294609build26-1777484162734001FjWO"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777488820140001xMpM","RegistrantKey":"1768245294609build26-1777488820140001xMpM"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/DEM320","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d2514ede-481a-4df9-97d9-2ea311457d23","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Cassidy Williams, Martin Woodward","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710246001rN7y","langLocale":"en-US","sessionId":"1774018710246001rN7y","sessionInstanceId":"1774018710246001rN7y","sessionCode":"DEM303","sessionCodeNormalized":"EN-US-DEM303","title":"Late to agentic coding? Don’t panic, build.","sortTitle":"late to agentic coding? don’t panic, build.","description":"​Feeling behind on agentic coding? You’re not. In this demo-heavy session, we’ll show how to go from idea to shipped. You’ll see practical patterns for planning tasks, delegating implementation, reviewing AI-generated PRs, and shipping with guardrails. No hype - just real workflows you can apply with your team right away.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Interactive Opening and Session Setup:</strong>  \nThe session begins with Martin Woodward and Cassidy Williams introducing themselves to the audience and encouraging participation <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:15\">00:00:15</a>. Attendees are invited to scan a QR code or visit a GitHub page to submit project ideas. They emphasize that the event will be hands-on, where they will build live applications using GitHub Copilot and discuss artificial intelligence–assisted coding. A quick survey of the audience reveals how many are using GitHub Copilot in different environments such as VS Code, Visual Studio, command line, and the recently announced GitHub app <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:40\">00:01:40</a>. After collecting these insights, the presenters navigate to the shared GitHub repository that houses instructions and demonstration resources <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>.</p>\n\n<p><strong>Demoing the GitHub Copilot App and Gathering Ideas:</strong>  \nCassidy demonstrates how the GitHub Copilot app works, showing that it is built on top of the Copilot CLI and SDKs while running the same engine locally <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. The audience is asked to suggest project ideas, and several fun examples are proposed—such as a whiskey review app, gesture-controlled ping pong, and a side-scrolling platformer game <a class=\"medius-aisummary\" data-starttime=\"00:04:09\">00:04:09</a>. Each suggestion automatically becomes an issue in their repository through Copilot’s integration, which even respects issue templates configured for idea creation. This demonstrates the intelligence of the app as it populates structured issues like “IDEA: Gesture-controlled game.” The presenters stress that Copilot can understand context, use templates, and perform tasks such as creating or managing issues in real time <a class=\"medius-aisummary\" data-starttime=\"00:11:24\">00:11:24</a>.</p>\n\n<p><strong>Industry Context and Developer Insights:</strong>  \nThe conversation shifts toward the rapid adoption of generative and agentic coding <a class=\"medius-aisummary\" data-starttime=\"00:05:11\">00:05:11</a>. Martin remarks that the pace of change in coding feels faster than the dot‑com boom, with pull requests authored by AI agents increasing dramatically. They highlight that GitHub itself relies on Copilot, which ranks among the top contributors to GitHub’s own codebase. Emphasizing the adoption curve, the hosts compare developers to early adopters and innovators, urging the audience to recognize their influence in shaping best practices while the majority of the industry still lags behind <a class=\"medius-aisummary\" data-starttime=\"00:06:54\">00:06:54</a>. They note that although 84% of developers at Microsoft reported using some form of AI, most non‑technical professionals have not, creating a distinct opportunity for this audience to lead in experimentation and organizational integration.</p>\n\n<p><strong>Live Build: Gesture-Controlled Rock‑Paper‑Scissors and Planning Mode:</strong>  \nReturning to live coding, they use Copilot’s assistant to research and prototype a new build app concept before moving on to implement the selected “Gesture-Controlled Rock‑Paper‑Scissors” game <a class=\"medius-aisummary\" data-starttime=\"00:13:31\">00:13:31</a>. The presenters explain different operational modes—interactive, plan, and autopilot (“Yolo mode”)—discussing why enterprises typically prefer guided planning for safety and oversight. Cassidy demonstrates “plan mode” with another audience suggestion, a side-scrolling platformer, showing model selection and reasoning levels across GPT and Claude families. They introduce “Rubber Duck,” a validation step where a second model reviews a plan for logic gaps, demonstrating measurable improvements from multi-model collaboration <a class=\"medius-aisummary\" data-starttime=\"00:19:05\">00:19:05</a>. Meanwhile, in the live test of their earlier game, Copilot builds and runs a browser-based version that successfully responds to face gestures—eliciting laughter and showcasing the tool’s efficiency <a class=\"medius-aisummary\" data-starttime=\"00:21:27\">00:21:27</a>.</p>\n\n<p><strong>Advanced CLI Features and Developer Guidance:</strong>  \nTo give developers deeper control, Martin walks through GitHub Copilot CLI commands and configuration <a class=\"medius-aisummary\" data-starttime=\"00:23:01\">00:23:01</a>. He explains the purpose of “Copilot Instructions” files, which encode a project’s standards and structure, and introduces “Spec Kit” for adding agent-friendly documentation. Cassidy then showcases features like remote control (“remote on”) for monitoring long tasks from a phone and “Chronicle” for automatic summaries, standups, or cost optimization. The “cost tips” feature is highlighted as a way to analyze model efficiency and token usage <a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>. Together, these tools demonstrate a growing ecosystem aimed at making generative development predictable, collaborative, and cost‑conscious.</p>\n\n<p><strong>Closing Thoughts and Call to Action:</strong>  \nAs the session closes, both presenters encourage the audience to continue experimenting with the tools introduced and visit the developer lab areas to collaborate further <a class=\"medius-aisummary\" data-starttime=\"00:25:46\">00:25:46</a>. Martin emphasizes that the key to mastering agentic coding is hands‑on practice, encouraging attendees to “just give it a go” in their own workflows. Cassidy thanks everyone for engaging, reaffirming that developers in the room are among the first shaping how AI‑assisted development will evolve. They conclude that the combination of Copilot, multi‑model orchestration, and community feedback is setting the foundation for a transformative future in software creation <a class=\"medius-aisummary\" data-starttime=\"00:26:04\">00:26:04</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","speakerIds":["1768245294609build26-1774284997876001pQKi","1768245294609build26-1774894273753001F5kk"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.788Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4c6da11e-f9a7-4141-b2ba-ffec4bfeb86e","ogTitle":"Late to agentic coding? Don’t panic, build.","ogDescription":"​Feeling behind on agentic coding? You’re not. In this demo-heavy session, we’ll show how to go from idea to shipped. You’ll see practical patterns for planning tasks, delegating implementation, reviewing AI-generated PRs, and shipping with guardrails. No hype - just real workflows you can apply with your team right away.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284997876001pQKi","RegistrantKey":"1768245294609build26-1774284997876001pQKi"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774894273753001F5kk","RegistrantKey":"1768245294609build26-1774894273753001F5kk"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM303","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"201065a1-f325-45dc-9dde-70c5516145a2","startDateTime":"2026-06-03T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roman Lutchyn","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779478233090001q3kx","langLocale":"en-US","sessionId":"1779478233090001q3kx","sessionInstanceId":"1779478233090001q3kx","sessionCode":"DEM314","sessionCodeNormalized":"EN-US-DEM314","title":"Majorana 2: The Topological Leap Toward Quantum Computing at Scale","sortTitle":"majorana 2: the topological leap toward quantum computing at scale","description":"Go beyond the keynote with Majorana 2 - our latest topological quantum chip, enabled by a reimagined material stack and AI‑accelerated R&D. Learn what one-minute qubit lifetimes mean for error correction, and discover the latest advances in the Microsoft Quantum software stack on the path to a scalable quantum machine.","aiDescription":"<p><strong>Introduction and the Promise of Quantum Computing:</strong> The speaker begins by highlighting the enormous transformative potential of quantum technologies for multiple industries (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>). Quantum computing, they note, can revolutionize healthcare by drastically reducing drug discovery and therapy development cycles, enhance energy and climate research through new catalysts for carbon capture and nitrogen fixation, and accelerate innovation across scientific domains. Microsoft aims to address this opportunity through a unified approach combining artificial intelligence (AI), quantum computation, and high-performance computing (HPC) to create hybrid workflows that are scalable and impactful (<a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>).</p>\n\n<p><strong>Microsoft’s Full-Stack Quantum Vision:</strong> At Microsoft, the focus is on developing a full-stack quantum system that integrates deeply from the hardware up through applications (<a class=\"medius-aisummary\" data-starttime=\"00:01:32\">00:01:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>). Rather than emphasizing qubit count, the company’s approach builds on every layer of the quantum stack, from physical processors and error correction methods to classical compute integration and software usability. The ultimate objective is for CPUs, GPUs, and QPUs to work seamlessly within Microsoft’s cloud environment to solve commercially valuable problems at scale (<a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a>).</p>\n\n<p><strong>Developing Reliable Topological Qubits:</strong> The presentation then dives into quantum hardware, where Microsoft’s research has centered on building small, fast, and reliable qubits (<a class=\"medius-aisummary\" data-starttime=\"00:02:34\">00:02:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:51\">00:02:51</a>). The company favors a topological approach to encoding quantum information, using Majorana modes—exotic quantum states that inherently protect against errors. Achieving this required inventing a new material stack and engineering a new phase of matter that supports these fault-tolerant states. The speaker proudly introduces “Majorana 2,” a new quantum processing unit based on these principles, integrating cryogenic CMOS controllers for efficient operation at extremely low temperatures (<a class=\"medius-aisummary\" data-starttime=\"00:03:29\">00:03:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:47\">00:04:47</a>).</p>\n\n<p><strong>The Science of New Quantum Materials:</strong> To achieve the reliability required for practical quantum computing, the team developed a hybrid phase of matter known as a topological superconductor, or “topoconductor” (<a class=\"medius-aisummary\" data-starttime=\"00:05:03\">00:05:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:31\">00:06:31</a>). This material emerges by precisely combining superconductors and semiconductors under specific magnetic conditions, enabling the manifestation of Majorana modes. The fabrication process involves atomic-scale engineering for pristine interfaces, a key factor for qubit performance. A major milestone in the research was replacing aluminum with lead as the superconducting layer, which has a fourfold larger energy gap, resulting in a thousandfold improvement in qubit performance over previous devices (<a class=\"medius-aisummary\" data-starttime=\"00:07:02\">00:07:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:04\">00:08:04</a>).</p>\n\n<p><strong>Scaling Qubit Arrays and Demonstrating Breakthroughs:</strong> The new materials enable Microsoft to design modular qubit arrays that can scale toward large, credit-card-sized chips containing millions of qubits (<a class=\"medius-aisummary\" data-starttime=\"00:09:22\">00:09:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:25\">00:10:25</a>). Lab results show devices achieving qubit lifetimes of up to one minute—an extraordinary feat compared to previous results of only 10 milliseconds (<a class=\"medius-aisummary\" data-starttime=\"00:12:02\">00:12:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:22\">00:12:22</a>). This thousandfold improvement validates Microsoft’s material-first design philosophy and underscores the effectiveness of AI-driven simulation and optimization tools. Based on these rapid advancements, Microsoft accelerated its roadmap for building a commercially viable quantum machine from 2033 to 2029 (<a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>).</p>\n\n<p><strong>AI-Accelerated Quantum Design and Conclusion:</strong> In closing, the speaker emphasizes how AI has dramatically shortened the design and optimization process for complex quantum chips—tasks that now take weeks instead of years (<a class=\"medius-aisummary\" data-starttime=\"00:13:33\">00:13:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:12\">00:14:12</a>). The combination of advanced materials, AI tools, and scalable architectures transforms what was once hard physics into a structured roadmap for a practical quantum computer. The talk concludes with a reaffirmation of Microsoft’s commitment to accelerating across every front—materials science, fabrication, algorithms, and architecture—to achieve quantum computing at scale, marking a confident step toward the future (<a class=\"medius-aisummary\" data-starttime=\"00:14:54\">00:14:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:05\">00:15:05</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:30 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/132d2247-cf6e-4b4d-92b1-e9040df49d5b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/132d2247-cf6e-4b4d-92b1-e9040df49d5b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/132d2247-cf6e-4b4d-92b1-e9040df49d5b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/132d2247-cf6e-4b4d-92b1-e9040df49d5b","speakerIds":["1768245294609build26-1780167366915001fBzx"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.236Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/132d2247-cf6e-4b4d-92b1-e9040df49d5b","ogTitle":"Majorana 2: The Topological Leap Toward Quantum Computing at Scale","ogDescription":"Go beyond the keynote with Majorana 2 - our latest topological quantum chip, enabled by a reimagined material stack and AI‑accelerated R&D. Learn what one-minute qubit lifetimes mean for error correction, and discover the latest advances in the Microsoft Quantum software stack on the path to a scalable quantum machine.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780167366915001fBzx","RegistrantKey":"1768245294609build26-1780167366915001fBzx"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"5dbedf22-c9a9-4015-a59b-667159c13d08","startDateTime":"2026-06-02T20:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vignesh Sridhar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774909657274001CJJw","langLocale":"en-US","sessionId":"1774909657274001CJJw","sessionInstanceId":"1774909657274001CJJw","sessionCode":"DEMSP383","sessionCodeNormalized":"EN-US-DEMSP383","title":"Move AI workflows from test to production on Microsoft Foundry","sortTitle":"move ai workflows from test to production on microsoft foundry","description":"Power use case-specific enterprise AI systems with high-performance inference from Fireworks AI integrated with Microsoft Foundry. In this live demo, see how teams move from test to production by running high‑performance inference directly on Foundry. Walk through an end‑to‑end workflow that shows how unified infrastructure improves latency, reduces cost, and simplifies deployment for real enterprise AI use cases.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video begins with Vignesh from Fireworks AI welcoming everyone to the session <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a> and explaining that he will present an overview of Fireworks AI, its capabilities, and how developers can use open-source models via Microsoft Foundry to build AI workflows <a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>. He introduces Fireworks AI as a platform for high-performance inference on open-source models, highlighting its founding team from PyTorch and Vertex AI and the platform’s day-zero support for leading open-source models like Kimi and GLM 5.1. The system handles massive scale—serving about 30 trillion tokens per day and 180,000 requests per second—while providing options for users to bring and deploy their own models through Foundry using the Fireworks serving stack optimized for low latency and high throughput <a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>.</p>\n\n<p><strong>Technical Capabilities and Azure Integration:</strong> Vignesh explains that Fireworks AI integrates natively with Azure, enabling developers to manage workloads directly on Foundry <a class=\"medius-aisummary\" data-starttime=\"00:01:08\">00:01:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:14\">00:01:14</a>. The Fireworks serving stack dynamically optimizes model performance using workload-aware techniques like adaptive caching and quantization and leverages their custom inference engine, Fire Retention, to balance latency and throughput requirements <a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>. Through abstraction, users are provided with production-ready endpoints for seamless deployment. On Foundry, Fireworks AI offers day-zero access to new open-source models that are optimized for enterprise-scale workloads, including support for custom weights and post-training registration workflows <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:49\">00:02:49</a>.</p>\n\n<p><strong>Use Cases and Model Comparison:</strong> The presentation highlights common industry use cases such as code completion, review bots, customized chatbots, and transcription or summarization workflows <a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:09\">00:03:09</a>. Vignesh shows how developers can experiment with multiple models through A/B testing on Foundry, selecting the best performing one to create and serve as an agent <a class=\"medius-aisummary\" data-starttime=\"00:03:16\">00:03:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>. Fireworks AI continuously adds new models to its portfolio, ensuring users can access the latest open-source innovations on launch day <a class=\"medius-aisummary\" data-starttime=\"00:03:35\">00:03:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:45\">00:03:45</a>. He then transitions to a live walkthrough on Microsoft Foundry, starting from the landing page and discovering Fireworks AI models available for immediate use <a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:28\">00:04:28</a>.</p>\n\n<p><strong>Deployment Workflow and Model Testing:</strong> The demo section explains how users can deploy models like Kimi K 2.6 for testing specific tasks <a class=\"medius-aisummary\" data-starttime=\"00:04:29\">00:04:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:03\">00:05:03</a>. Multi-tenant serverless endpoints allow quick testing and evaluation before moving to dedicated single-tenant deployments optimized through provision throughput calculations that determine required resources for production-scale workloads <a class=\"medius-aisummary\" data-starttime=\"00:05:27\">00:05:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>. He demonstrates testing models in a playground, comparing performance and latency between deployed models like Minimax and Kimi 2.6, showing how developers can interact with each model, inspect response quality, and compare outputs side by side to select the most suitable one for their use case <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>.</p>\n\n<p><strong>Evaluation and Production Deployment:</strong> After selecting the best model, users can convert it into an agent and evaluate its performance using Foundry’s matrix configurator <a class=\"medius-aisummary\" data-starttime=\"00:08:42\">00:08:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:39\">00:09:39</a>. Vignesh demonstrates configuring evaluation metrics such as intent resolution, coherence, fluency, and relevance for datasets with ground truth comparisons <a class=\"medius-aisummary\" data-starttime=\"00:09:49\">00:09:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:54\">00:10:54</a>. He reviews completed evaluations showing metric scores and explains how developers can iterate and refine their agents for improved performance prior to production <a class=\"medius-aisummary\" data-starttime=\"00:11:01\">00:11:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:28\">00:11:28</a>. Once optimized, users can publish agents either as Foundry web apps or integrate them directly into code projects using API endpoints and Python snippets <a class=\"medius-aisummary\" data-starttime=\"00:11:34\">00:11:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:14\">00:12:14</a>.</p>\n\n<p><strong>Fine-Tuning and Closing Remarks:</strong> In closing, Vignesh explains advanced workflows for fine-tuning where developers can bring their own weights or base models to Fireworks AI via Foundry <a class=\"medius-aisummary\" data-starttime=\"00:12:44\">00:12:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:27\">00:13:27</a>. Using frameworks like SFT or RFT, models can be customized for specific use cases and redeployed through the same inference-serving workflow. He concludes by thanking the audience and inviting questions <a class=\"medius-aisummary\" data-starttime=\"00:13:32\">00:13:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:41\">00:13:41</a>. The session ends with Manesh expressing gratitude for the presentation and sharing a brief announcement about the next upcoming session, encouraging participants to stay or have a great day <a class=\"medius-aisummary\" data-starttime=\"00:13:58\">00:13:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:26\">00:14:26</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"20:40 - 21:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:05:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/2d0d8c96-5737-41f7-8a54-7be68b88101f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/2d0d8c96-5737-41f7-8a54-7be68b88101f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/2d0d8c96-5737-41f7-8a54-7be68b88101f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/2d0d8c96-5737-41f7-8a54-7be68b88101f","speakerIds":["1768245294609build26-1776705717450001Bz8g"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.843Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Fireworks AI"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/2d0d8c96-5737-41f7-8a54-7be68b88101f","ogTitle":"Move AI workflows from test to production on Microsoft Foundry","ogDescription":"Power use case-specific enterprise AI systems with high-performance inference from Fireworks AI integrated with Microsoft Foundry. In this live demo, see how teams move from test to production by running high‑performance inference directly on Foundry. Walk through an end‑to‑end workflow that shows how unified infrastructure improves latency, reduces cost, and simplifies deployment for real enterprise AI use cases.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776705717450001Bz8g","RegistrantKey":"1768245294609build26-1776705717450001Bz8g"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/DEMSP383%20-%20MSFT%20Build%20-%20Demo%20Slides_1779148454976001wZDS_1780513247374001bkpY.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"}]},{"@search.score":1.0,"documentId":"690a439c-a175-4e67-8a89-d23c8b5e8430","startDateTime":"2026-06-03T17:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jan Kalis, Vini Soto","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018706129001r866","langLocale":"en-US","sessionId":"1774018706129001r866","sessionInstanceId":"1774018706129001r866","sessionCode":"DEM312","sessionCodeNormalized":"EN-US-DEM312","title":"Multi-agents in action with 3 AI agents, 3 frameworks, tools & models","sortTitle":"multi-agents in action with 3 ai agents, 3 frameworks, tools & models","description":"Three agent frameworks. One Agentic Startup Content Factory. Zero manual steps. In this live demo we build an intelligent system that researches topics, writes articles and code samples, and refines its own output, all while running autonomously. We'll deploy LangGraph, .NET Microsoft Agent Framework, and GitHub Copilot SDK agents to Azure Container Apps, then connect Microsoft Foundry for observability and evals. You'll walk out with an Agentic Startup Content Factory you can clone and ship.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The session opens with an introduction by Jan and Vinnie from the Azure Container Apps team. They welcome the audience to their presentation on multi-agents in action and outline the plan to showcase three different agent frameworks running at scale (<a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>). The agenda includes explaining what modern agentic infrastructure is, why it matters, and demonstrating the \"Agentic Content Factory\" demo. They also reference a forthcoming GitHub repository link for attendees to explore after the talk (<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>). Gartner’s forecast that 40% of agentic projects will face cancellations by 2025 is highlighted, attributed not to AI model limitations but to issues like governance, runtime instability, and unmonitored budgets (<a class=\"medius-aisummary\" data-starttime=\"00:00:57\">00:00:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>).</p>\n\n<p><strong>Introducing Azure Container App Sandboxes:</strong> The speakers discuss typical runtime problems—like cold starts, untrusted code execution, and lack of lifecycle control—and introduce Azure Container App Sandboxes as a solution (<a class=\"medius-aisummary\" data-starttime=\"00:02:28\">00:02:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:56\">00:02:56</a>). They announce a private preview of this feature, emphasizing its secure, isolated, and stateful infrastructure on demand. Users can run code safely with snapshot-based restoration and near-instant resumption (<a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>). Internal teams like Microsoft Foundry and GitHub are already leveraging this for hosted agents. They also showcase Azure Container Apps Express, described as an agent-first technology that offers provisioning in seconds and near-zero scale-up latency (<a class=\"medius-aisummary\" data-starttime=\"00:03:43\">00:03:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:57\">00:03:57</a>).</p>\n\n<p><strong>Three-Agent Architecture and System Design:</strong> The demo portion introduces three collaborating agents—a researcher, a content creator, and a narrator—working together within the content factory pipeline (<a class=\"medius-aisummary\" data-starttime=\"00:04:02\">00:04:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>). Each agent uses a different technology stack: LangGraph in Python for research, Agent Framework in C#, and GitHub Copilot SDK for narrative generation. Managed through a Foundry orchestrator, these agents emit OpenTelemetry (Otel) data to allow system-wide observability. The goal is to demonstrate how multi-agent pipelines can operate under one management plane, regardless of framework diversity. Agent telemetry and metrics are collected across all applications to illustrate real-time coordination and monitoring (<a class=\"medius-aisummary\" data-starttime=\"00:06:03\">00:06:03</a>).</p>\n\n<p><strong>Live Demo: Soccer Simulation and Sandboxes in Action:</strong> For an engaging use case, the presenters run a soccer match simulation between Mexico and the Czech Republic, dynamically executed by the agents through sandbox environments (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:02\">00:10:02</a>). The research agent fetches live data, executes dynamic Python code in isolated sandboxes, and leverages Foundry’s LLMs to predict outcomes. Each sandbox represents a temporary, secure compute environment that spins up, executes code, then snapshots and idles automatically when inactive (<a class=\"medius-aisummary\" data-starttime=\"00:11:20\">00:11:20</a>). They demonstrate security through egress and access policies—showing how URLs can be blocked, outbound calls controlled, and secrets protected from exposure by injecting API keys via an external egress gateway rather than hardcoding them inside agent logic (<a class=\"medius-aisummary\" data-starttime=\"00:16:02\">00:16:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:06\">00:17:06</a>).</p>\n\n<p><strong>Security, Observability, and Workflow Management:</strong> The team explains secure credential handling using Azure Managed Identity for internal API access, alongside transformation-based egress policies that transparently attach secure keys to outbound requests (<a class=\"medius-aisummary\" data-starttime=\"00:14:09\">00:14:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:49\">00:16:49</a>). Observability is achieved through unified OpenTelemetry aggregation into Application Insights, allowing administrators to trace agent flows, performance metrics, and errors (<a class=\"medius-aisummary\" data-starttime=\"00:18:40\">00:18:40</a>). The demo also shows continuous evaluation and centralized monitoring through Microsoft Foundry, regardless of whether agents use different SDKs or programming languages. This unified oversight helps maintain governance while managing large multi-agent systems (<a class=\"medius-aisummary\" data-starttime=\"00:19:19\">00:19:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:28\">00:19:28</a>).</p>\n\n<p><strong>Conclusion and Next Steps:</strong> Jan and Vinnie conclude by recapping that Azure Container App Sandboxes enable fast, stateful, and isolated execution for multi-agent workloads (<a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:12\">00:20:12</a>). They emphasize security-by-design, observability, and cost efficiency achieved via automated idling and on-demand scaling. Attendees are encouraged to visit the Azure Container Apps booth and to try sandboxes directly at sandboxes.azure.com (<a class=\"medius-aisummary\" data-starttime=\"00:21:10\">00:21:10</a>). The presenters close with an invitation to upcoming sessions and demos at Build, thanking the audience for participating and reiterating Microsoft’s mission to make multi-agent orchestration secure, manageable, and developer-friendly.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:00 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/d23cefd0-8854-4932-a055-1022d8c5aa15","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d23cefd0-8854-4932-a055-1022d8c5aa15","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d23cefd0-8854-4932-a055-1022d8c5aa15","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d23cefd0-8854-4932-a055-1022d8c5aa15","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d23cefd0-8854-4932-a055-1022d8c5aa15","speakerIds":["1768245294609build26-1769104859708001i6bQ","1768245294609build26-1775784866500001O4MP"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T23:36:49.777Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d23cefd0-8854-4932-a055-1022d8c5aa15","ogTitle":"Multi-agents in action with 3 AI agents, 3 frameworks, tools & models","ogDescription":"Three agent frameworks. One Agentic Startup Content Factory. Zero manual steps. In this live demo we build an intelligent system that researches topics, writes articles and code samples, and refines its own output, all while running autonomously. We'll deploy LangGraph, .NET Microsoft Agent Framework, and GitHub Copilot SDK agents to Azure Container Apps, then connect Microsoft Foundry for observability and evals. You'll walk out with an Agentic Startup Content Factory you can clone and ship.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775784866500001O4MP","RegistrantKey":"1768245294609build26-1775784866500001O4MP"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769104859708001i6bQ","RegistrantKey":"1768245294609build26-1769104859708001i6bQ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM312","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"7cfb0da8-07f7-44ba-84f9-acacd723d120","startDateTime":"2026-06-03T17:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prachi Sethi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779487223382001nlXF","langLocale":"en-US","sessionId":"1779487223382001nlXF","sessionInstanceId":"1779487223382001nlXF","sessionCode":"DEM306","sessionCodeNormalized":"EN-US-DEM306","title":"OM1: Modular by Design, Hardware-Agnostic by Default. Your gateway to next-gen robotics","sortTitle":"om1: modular by design, hardware-agnostic by default. your gateway to next-gen robotics","description":"Robots today are powerful but siloed. Each one is locked to its own stack, hardware, and interfaces. OM1 is a modular, hardware-agnostic orchestration layer that bridges any robot with the cognitive infrastructure needed to make it actually useful. In this session, we'll walk through OM1's architecture, show how its modular design allows you to swap or add capabilities, and get started on your laptops. We'll also demo it live on the Unitree Go2 and in a cloud simulator. Bring a laptop if you'd like to follow along!","aiDescription":"<p><strong>Introduction and Agenda Overview:</strong> The presentation begins with a friendly greeting and an introduction to Open Mind and their product, OM1 (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>). The speaker outlines the agenda, which includes walking through OM1’s architecture, setup instructions, contribution guidelines, and a live demonstration involving a robotic dog and the Cloud Axiom platform (<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>). OM1, an open-source, hardware-agnostic cognitive layer, connects a wide variety of robots to an AI-driven orchestration infrastructure. It provides social intelligence to robots and currently supports platforms like Unitree, Unitary G1, Limuxtron, and Turtlebot, with flexible deployment across Mac and Ubuntu systems (<a class=\"medius-aisummary\" data-starttime=\"00:00:42\">00:00:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>).</p>\n\n<p><strong>Architecture and Intelligence Features:</strong> The speaker explains the modular design of OM1, where each robot connects through a unified interface regardless of hardware differences in sensors, microphones, GPS, and other components (<a class=\"medius-aisummary\" data-starttime=\"00:01:24\">00:01:24</a>). The system processes environmental inputs and user prompts such as movement or personality instructions that define how a robot behaves or speaks. For instance, the robot “Bytes” recognizes its name and reacts based on language models and personality prompts configured by users (<a class=\"medius-aisummary\" data-starttime=\"00:01:59\">00:01:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>). OM1 integrates with large language models (LLMs) to interpret user instructions, enabling robots to perform diverse tasks like speaking, moving, or reporting system metrics. It supports autonomous navigation, SLAM map generation, obstacle avoidance, face detection, and even auto-charging for select units like Unitree Go2 (<a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>).</p>\n\n<p><strong>Open Source Collaboration and Cloud Setup:</strong> The project being open source encourages developers to contribute new robot drivers and sensor support through GitHub (<a class=\"medius-aisummary\" data-starttime=\"00:03:03\">00:03:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>). The demonstration transitions to a practical guide showing how to clone the repository, install dependencies, and set up API keys via OM1’s cloud portal (<a class=\"medius-aisummary\" data-starttime=\"00:07:47\">00:07:47</a>). It includes defining a robot’s configuration in YAML with details like version, API key, and system prompts that shape the robot’s personality—whether amusing, professional, or supportive (<a class=\"medius-aisummary\" data-starttime=\"00:10:01\">00:10:01</a>). The session also demonstrates integrating microphones, cameras, and audio through Google ASR, visual perception through BLIP or VILA, and LLM processing options that include OpenAI, Gemini, Anthropic, or locally hosted models using Ollama (<a class=\"medius-aisummary\" data-starttime=\"00:11:13\">00:11:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:33\">00:11:33</a>).</p>\n\n<p><strong>Live Demonstration and Go Migration:</strong> The first demo displays “Bytes” the robotic dog responding to commands while connected to a laptop and cloud simulator (<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>). The system runs virtual agents that handle voice and text input through containerized environments, highlighting OM1’s dockerized deployment flexibility. The speaker introduces a Go-language implementation of OM1 that replaces the earlier Python version, improving performance and compatibility with operating systems such as Windows, Linux, and Raspberry Pi setups (<a class=\"medius-aisummary\" data-starttime=\"00:12:46\">00:12:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:07\">00:13:07</a>). The Go migration showcases reduced latency and ease of binary-based deployment, supporting local and remote robot control environments.</p>\n\n<p><strong>Simulation, Mapping, and Navigation Features:</strong> The cloud simulator allows users to observe live robot streams and control navigation in simulated apartments or warehouses (<a class=\"medius-aisummary\" data-starttime=\"00:13:30\">00:13:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>). During the demo, the robot autonomously maps its environment via SLAM and recognizes predefined locations like “kitchen” or “living room.” In navigation mode, verbal commands guide robots to those saved coordinates. Additional interactive modes like greeting or person-following enable robots to detect individuals, approach them, and engage in brief contextual exchanges (<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:11\">00:16:11</a>). The App Store functionality lets developers upload customized robot personalities or tasks, linked through each device’s API key configuration (<a class=\"medius-aisummary\" data-starttime=\"00:17:02\">00:17:02</a>).</p>\n\n<p><strong>Telepresence Capability and Closing Remarks:</strong> The final portion of the session highlights telepresence, allowing remote video communication through robots for practical scenarios like checking on family members or home spaces (<a class=\"medius-aisummary\" data-starttime=\"00:17:36\">00:17:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:34\">00:18:34</a>). Users can navigate their robots remotely, access live camera feeds, and initiate video calls, enhancing OM1’s value for care and monitoring applications. The demonstration concludes with saving maps, switching between autonomous modes, and exploring remote teleoperation through the simulator. The speaker wraps up by inviting questions and encouraging contributions to the Open Mind ecosystem before ending the demo session (<a class=\"medius-aisummary\" data-starttime=\"00:20:40\">00:20:40</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"17:00 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/f9818341-9d02-497c-9b46-cb5f25533179","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/f9818341-9d02-497c-9b46-cb5f25533179","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/f9818341-9d02-497c-9b46-cb5f25533179","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/f9818341-9d02-497c-9b46-cb5f25533179","speakerIds":["1768245294609build26-1779487295106001adIc"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.237Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/f9818341-9d02-497c-9b46-cb5f25533179","ogTitle":"OM1: Modular by Design, Hardware-Agnostic by Default. Your gateway to next-gen robotics","ogDescription":"Robots today are powerful but siloed. Each one is locked to its own stack, hardware, and interfaces. OM1 is a modular, hardware-agnostic orchestration layer that bridges any robot with the cognitive infrastructure needed to make it actually useful. In this session, we'll walk through OM1's architecture, show how its modular design allows you to swap or add capabilities, and get started on your laptops. We'll also demo it live on the Unitree Go2 and in a cloud simulator. Bring a laptop if you'd like to follow along!","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779487295106001adIc","RegistrantKey":"1768245294609build26-1779487295106001adIc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"OSS","logicalValue":"OSS"}]},{"@search.score":1.0,"documentId":"04e23243-7f64-4986-be48-c5010b3e01ea","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Michel Hubert","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290371737001Dgv0","langLocale":"en-US","sessionId":"1776290371737001Dgv0","sessionInstanceId":"1776290371737001Dgv0","sessionCode":"DEM367","sessionCodeNormalized":"EN-US-DEM367","title":"Onboarding a Developer with AI: A Better First-Day Experience","sortTitle":"onboarding a developer with ai: a better first-day experience","description":"Onboarding is one of the most underestimated areas of Developer Experience. This session demonstrates how AI can help a new developer become productive faster by explaining a codebase, surfacing useful context, identifying key components, and supporting a first contribution. The demo shows how AI can reduce frustration and shorten the time to meaningful impact.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a>, Michel Riber greets the audience and introduces the central theme—how developer onboarding transforms when supported by AI agents. He clarifies that the session is focused not on HR or administrative onboarding but on the moment a new developer encounters an unfamiliar repository and wonders where to begin. Michel, an MVP of 17 years representing Avanade France, shares statistics at <a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a> to establish the problem: it takes 3–9 months for developers to become fully productive, 40% of the first week is spent merely reading code, and one in three questions newcomers have go unasked. He emphasizes that this is not a human resources challenge but a matter of developer experience.</p>\n\n<p><strong>The Four \"Actors\" Framework:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:02:27\">00:02:27</a>, Michel introduces a four-part structure representing stages of an AI-enhanced onboarding process: understanding the codebase rapidly, uncovering hidden context, identifying key components and contributors, and shipping a valid first pull request. He explains that the demonstration will follow a fictional developer named Sara through each of these stages, using a real-world open source project as an example. Michel notes the immense challenge Sara faces on day one—more than 2,000 source files spanning five languages and outdated documentation (<a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:40\">00:04:40</a>). The objective is for Sara to produce her first PR by the end of the week, an implicit expectation among teams.</p>\n\n<p><strong>Understanding and Mapping the Codebase:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a> to <a class=\"medius-aisummary\" data-starttime=\"00:08:55\">00:08:55</a>, the demo shows Sara’s AI agent generating a full architectural comprehension of the repository. She prompts the agent to “give me a tour” of the system—its entry points and data flow—and within minutes receives what would normally require hours of explanation from a tech lead. The AI analyzes code connectivity, recent commits, and boundaries to describe the application structure—from React front-end to Node.js back-end—and highlights active versus legacy components. It even constructs diagrams of classes and dependencies automatically by reading the code, not documentation. Sara now gains a visual, functional understanding of the entire system, including insights into which components have recently changed or remain dormant.</p>\n\n<p><strong>Surfacing Context and Team Conventions:</strong> Moving to <a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>, Michel demonstrates how Sara’s agent extracts historical reasoning behind specific code modules, such as why a file was created and how certain technologies evolved in the project. By parsing pull request descriptions, commits, and issue threads, the AI reconstructs the decision-making timeline—revealing, for instance, the project's migration from PostgreSQL to ClickHouse for performance reasons (<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:17\">00:13:17</a>). The agent then analyzes unwritten team conventions for logging, error handling, and naming by examining live code rather than outdated manuals (<a class=\"medius-aisummary\" data-starttime=\"00:14:02\">00:14:02</a>). Sara now learns not only what the project does but also the “why” behind it and how to adhere to team practices from the first contribution.</p>\n\n<p><strong>Prioritization and Performing the First PR:</strong> During <a class=\"medius-aisummary\" data-starttime=\"00:15:06\">00:15:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:54\">00:19:54</a>, the AI helps Sara prioritize key files to study, identify core contributors, and finally tackle an actual issue in the open source project. The agent lists the top relevant files, spots module owners through commit history, and assists in building a plan for code modification. When Sara selects a real issue (ID 58757), the agent reads it, identifies the required changes, provides a strategy without altering code, then implements the fix and runs regression tests. Once validated, it even drafts a pull request summary aligning with the repository’s stylistic conventions. The automation covers technical updates, validation, and documentation synthesis—all in a structured, transparent flow guided by context-aware reasoning.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> Finally, at <a class=\"medius-aisummary\" data-starttime=\"00:22:55\">00:22:55</a> Michel concludes that onboarding is no longer a test of patience but of curiosity. With AI as a contextual engine rather than merely a code generator, developer onboarding shifts toward comprehension instead of repetition. He highlights metrics such as “time to first PR” as essential for evaluating AI’s impact and encourages teams to compare pre- and post-AI onboarding speeds. Michel reminds that the agent complements, not replaces, human collaboration—serving as a programming partner that restores focus to learning and problem-solving (<a class=\"medius-aisummary\" data-starttime=\"00:23:00\">00:23:00</a>). He closes by inviting questions via LinkedIn and thanking viewers while wishing them a productive build event (<a class=\"medius-aisummary\" data-starttime=\"00:23:42\">00:23:42</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:50 - 19:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:15:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/3f96eb60-e022-4ac1-af17-13b71128c648","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/3f96eb60-e022-4ac1-af17-13b71128c648","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/3f96eb60-e022-4ac1-af17-13b71128c648","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/3f96eb60-e022-4ac1-af17-13b71128c648","speakerIds":["1768245294609build26-1757452741184001uVMn"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.904Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/3f96eb60-e022-4ac1-af17-13b71128c648","ogTitle":"Onboarding a Developer with AI: A Better First-Day Experience","ogDescription":"Onboarding is one of the most underestimated areas of Developer Experience. This session demonstrates how AI can help a new developer become productive faster by explaining a codebase, surfacing useful context, identifying key components, and supporting a first contribution. The demo shows how AI can reduce frustration and shorten the time to meaningful impact.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1757452741184001uVMn","RegistrantKey":"1768245294609build26-1757452741184001uVMn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"adcdffaf-c9b3-4f61-b772-aeb32cd16684","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rachel Baskin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774632913980001rhde","langLocale":"en-US","sessionId":"1774632913980001rhde","sessionInstanceId":"1774632913980001rhde","sessionCode":"DEMSP393","sessionCodeNormalized":"EN-US-DEMSP393","title":"Orchestrate omnichannel AI agents with Foundry and Twilio","sortTitle":"orchestrate omnichannel ai agents with foundry and twilio","description":"Customer conversations often lose context across channels, making it difficult to build cohesive AI agents. Learn how to close the gap with Microsoft Foundry and Twilio. Use Foundry to establish models, voice, and agents on Azure. Add Twilio for cross-channel memory, real-time orchestration, and conversation intelligence, all wired by the Agent Connect SDK. See a demo and learn patterns for building scalable, real-time workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The video begins with Rachel, a product manager at Twilio, greeting the audience and introducing the topic of building omnichannel AI agents using Twilio and Microsoft Foundry <a class=\"medius-aisummary\" data-starttime=\"00:00:09\">00:00:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>. She highlights the rising global investment in AI, noting Gartner’s reported 44% year-over-year increase. Despite this growth, over half of surveyed leaders report getting little value from AI, and only 6% of brands have improved customer experience scores. Rachel points out this gap between AI spending and actual ROI, setting the stage for why Twilio and Foundry are working together to enable more practical, integrated AI solutions <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>.</p>\n\n<p><strong>Challenges in AI Agent Development:</strong> Rachel explains that the biggest challenge companies face is “memory loss” in customer interaction systems <a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>. Traditional support systems forget customer history across channels, forcing users to repeat information and causing poor experiences. She emphasizes how this lack of persistent context affects both virtual and human agents. Additionally, she notes that deploying omnichannel AI agents capable of maintaining context, memory, and functioning reliably at scale is technically complex <a class=\"medius-aisummary\" data-starttime=\"00:02:15\">00:02:15</a>. These difficulties inspired Twilio to collaborate with Microsoft Foundry and develop Twilio Agent Connect to simplify building and deploying enterprise-ready agents <a class=\"medius-aisummary\" data-starttime=\"00:02:48\">00:02:48</a>.</p>\n\n<p><strong>Introducing Twilio Agent Connect and Foundry Integration:</strong> Rachel describes the integration between Twilio and Microsoft Foundry as a way to unify Twilio’s communication layer with Foundry’s intelligence models <a class=\"medius-aisummary\" data-starttime=\"00:03:43\">00:03:43</a>. Twilio powers the real-time conversation management and data continuity, while Foundry provides model governance, optimization, and reasoning through its AI tools. Twilio Agent Connect acts as the open-source SDK that bridges them, offering a simpler path from concept to production with fewer engineering overheads <a class=\"medius-aisummary\" data-starttime=\"00:04:05\">00:04:05</a>. This collaboration ensures the creation of a fully functional omnichannel AI solution where every communication—voice, text, or chat—shares persistent customer memory across sessions.</p>\n\n<p><strong>Architectural Overview and Hosted Agents:</strong> Initially, Twilio Agent Connect used container-based deployments via Azure services <a class=\"medius-aisummary\" data-starttime=\"00:04:19\">00:04:19</a>. Rachel details how container apps allowed developers full control but were not ideal for long-lived, per-user sessions typical of AI agents. To solve scalability and security isolation issues, Microsoft introduced “hosted agents,” a new compute model built for persistent, secure, stateful sessions <a class=\"medius-aisummary\" data-starttime=\"00:05:22\">00:05:22</a>. Hosted agents come with automatic sandboxing, predictable scaling, cost-efficiency (including scale-to-zero), and enterprise-level monitoring and evaluation tools. This new serverless architecture eliminates infrastructure management while maintaining performance for intelligent conversational systems <a class=\"medius-aisummary\" data-starttime=\"00:07:20\">00:07:20</a>.</p>\n\n<p><strong>Live Demo: Deploying and Connecting a Voice and SMS Agent:</strong> Rachel transitions to a live demonstration <a class=\"medius-aisummary\" data-starttime=\"00:08:30\">00:08:30</a>, walking through deploying Twilio Agent Connect with Azure hosted agents using the open-source GitHub repository. She shows how deploying requires minimal setup via a single command that provisions resources automatically. Then she illustrates how Twilio phone numbers and messaging conversations are connected to webhooks within the API management service, linking them to the AI agent <a class=\"medius-aisummary\" data-starttime=\"00:11:05\">00:11:05</a>. In the demo, she conducts a simulated airline customer interaction: a phone call where the AI instantly recognizes the customer’s flight and updates a seat <a class=\"medius-aisummary\" data-starttime=\"00:13:01\">00:13:01</a>. Later, she follows up by texting the same number to add a checked bag, demonstrating that the system remembers previous interactions and context seamlessly across channels <a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>.</p>\n\n<p><strong>Q&A and Closing Remarks:</strong> In the final segment, Rachel answers audience questions about extending business logic, customizing prompts, and using Twilio Conversation Relay versus the new native integration <a class=\"medius-aisummary\" data-starttime=\"00:17:16\">00:17:16</a>. She explains that developers can modify agent behavior, connect custom knowledge bases, and choose models directly through the hosted agent environment or GitHub examples. The discussion also touches on deployment strategies for multi-tenant scenarios and improvements enabled by direct websocket connections for faster voice responsiveness <a class=\"medius-aisummary\" data-starttime=\"00:19:25\">00:19:25</a>. Rachel concludes by encouraging viewers to explore the open-source repository for hands-on experimentation and further learning, emphasizing that this integration marks a step forward in making enterprise-grade omnichannel AI systems easier to deploy and scale <a class=\"medius-aisummary\" data-starttime=\"00:21:10\">00:21:10</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c1269a04-04cc-4d61-b461-41f96f288bac","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c1269a04-04cc-4d61-b461-41f96f288bac","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c1269a04-04cc-4d61-b461-41f96f288bac","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c1269a04-04cc-4d61-b461-41f96f288bac","speakerIds":["1768245294609build26-1778617521151001cKpm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.842Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Twilio"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c1269a04-04cc-4d61-b461-41f96f288bac","ogTitle":"Orchestrate omnichannel AI agents with Foundry and Twilio","ogDescription":"Customer conversations often lose context across channels, making it difficult to build cohesive AI agents. Learn how to close the gap with Microsoft Foundry and Twilio. Use Foundry to establish models, voice, and agents on Azure. Add Twilio for cross-channel memory, real-time orchestration, and conversation intelligence, all wired by the Agent Connect SDK. See a demo and learn patterns for building scalable, real-time workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778617521151001cKpm","RegistrantKey":"1768245294609build26-1778617521151001cKpm"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Twilio_Orchestrate%20Omnichannel%20AI%20Agents_1780095219631001R1T0_1780513348700001nbCe.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Software Development Company","logicalValue":"Software Development Company"},{"displayValue":"Foundry Control Plane","logicalValue":"Foundry Control Plane"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"Personalization","logicalValue":"Personalization"},{"displayValue":"App Integration","logicalValue":"App Integration"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"ceeec8b3-8e1a-4bfd-842a-d44e4dc43321","startDateTime":"2026-06-03T18:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vijay Aski, Chris Lauren","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780457275981001o2ao","langLocale":"en-US","sessionId":"1780457275981001o2ao","sessionInstanceId":"1780457275981001o2ao","sessionCode":"DEM321-R1","sessionCodeNormalized":"EN-US-DEM321-R1","title":"Post-Training and Deploying Open Source Reasoning Models in Foundry","sortTitle":"post-training and deploying open source reasoning models in foundry","description":"Go beyond prompt engineering to build custom reasoning models using reinforcement learning in Microsoft Foundry. We'll walk through how to train and fine-tune a model to improve reasoning quality, deploy it into Foundry, and integrate it into an agent workflow. Designed for developers comfortable with code, this session focuses on real implementation details, covering training loops, evaluation, and deployment patterns that directly impact agent performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:20 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0ffd6080-1f08-4054-9eea-414215ebfd07","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0ffd6080-1f08-4054-9eea-414215ebfd07","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0ffd6080-1f08-4054-9eea-414215ebfd07","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0ffd6080-1f08-4054-9eea-414215ebfd07","speakerIds":["1768245294609build26-1776895269615001aQWT","1768245294609build26-1776456842448001LPgI"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.465Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0ffd6080-1f08-4054-9eea-414215ebfd07","ogTitle":"Post-Training and Deploying Open Source Reasoning Models in Foundry","ogDescription":"Go beyond prompt engineering to build custom reasoning models using reinforcement learning in Microsoft Foundry. We'll walk through how to train and fine-tune a model to improve reasoning quality, deploy it into Foundry, and integrate it into an agent workflow. Designed for developers comfortable with code, this session focuses on real implementation details, covering training loops, evaluation, and deployment patterns that directly impact agent performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776456842448001LPgI","RegistrantKey":"1768245294609build26-1776456842448001LPgI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776895269615001aQWT","RegistrantKey":"1768245294609build26-1776895269615001aQWT"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"99693808-e03a-4982-aad3-b60da8c218db","startDateTime":"2026-06-02T21:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vijay Aski, Chris Lauren","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018714055001rncm","langLocale":"en-US","sessionId":"1774018714055001rncm","sessionInstanceId":"1774018714055001rncm","sessionCode":"DEM321","sessionCodeNormalized":"EN-US-DEM321","title":"Post-Training and Deploying Open Source Reasoning Models in Foundry","sortTitle":"post-training and deploying open source reasoning models in foundry","description":"Go beyond prompt engineering to build custom reasoning models using reinforcement learning in Microsoft Foundry. We'll walk through how to train and fine-tune a model to improve reasoning quality, deploy it into Foundry, and integrate it into an agent workflow. Designed for developers comfortable with code, this session focuses on real implementation details, covering training loops, evaluation, and deployment patterns that directly impact agent performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session begins with light humor and an apology for a technical delay before moving into the core topic of intelligent agents <a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>. The speaker highlights that while agents are gaining popularity for handling autonomous workflows, they introduce scalability challenges due to high token consumption <a class=\"medius-aisummary\" data-starttime=\"00:00:32\">00:00:32</a>. The discussion sets the stage by comparing agents with traditional chat scenarios, emphasizing the need for cost-effective ways to deploy them in production through optimized model usage.</p>\n\n<p><strong>Optimizing Agent Efficiency with Foundry:</strong> To address cost and performance issues, the presenter introduces Foundry as a solution for fine-tuning smaller, open-source models that can outperform larger frontier models once customized for a specific business domain <a class=\"medius-aisummary\" data-starttime=\"00:00:56\">00:00:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>. Foundry allows users to define what “good” looks like in their workflows and iteratively train models at a fraction of the cost. The speaker stresses that rather than relying solely on massive proprietary models, organizations can leverage Foundry to control their data, optimize their agent logic, and achieve scalable, efficient performance <a class=\"medius-aisummary\" data-starttime=\"00:01:56\">00:01:56</a>.</p>\n\n<p><strong>Test-Driven Development for Agents and Evaluation Framework:</strong> A major conceptual shift is proposed—treating agent evaluation similarly to software testing. By defining evaluation criteria, or “evals”, upfront, teams can iteratively test and refine their models <a class=\"medius-aisummary\" data-starttime=\"00:02:09\">00:02:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>. The speaker parallels this approach with test-driven development, explaining that these evaluations govern expected behaviors, policy adherence, and tool usage patterns. Foundry now supports creating, deploying, and customizing agents using any open-source framework, ensuring that organizations can maintain consistency, trust, and quality regardless of model source <a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:25\">00:03:25</a>.</p>\n\n<p><strong>Demonstration of Model Customization in Foundry:</strong> The live demonstration begins with a practical setup of Foundry’s workspace <a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>. The speaker contrasts the performance of the GPT 5.2 frontier model and an open-source model, Quen 314B, within a customer service agent scenario <a class=\"medius-aisummary\" data-starttime=\"00:03:49\">00:03:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:43\">00:04:43</a>. While GPT 5.2 performs well but at a higher cost, the smaller Quen model demonstrates significant efficiency improvements when fine-tuned using reinforcement learning techniques <a class=\"medius-aisummary\" data-starttime=\"00:05:07\">00:05:07</a>. This process teaches the model not only to understand the business language but also to properly sequence tool interactions to achieve faster resolutions at lower costs.</p>\n\n<p><strong>Automated Data Collection and Reinforcement Learning Cycle:</strong> The presentation continues with an explanation of Foundry’s built-in observability and dataset creation features. Every user session and model interaction is traced and stored automatically, allowing teams to audit model behavior and cost <a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:06\">00:06:06</a>. These traces can then be transformed into structured training datasets that feed future reinforcement learning rounds, continuously improving model effectiveness over time <a class=\"medius-aisummary\" data-starttime=\"00:06:30\">00:06:30</a>. This closed-loop approach ensures steady enhancement of accuracy and alignment between production data and model logic.</p>\n\n<p><strong>Deep Dive into Post-Training Techniques and Conclusion:</strong> Vijay transitions into the technical segment, explaining the concept of post-training as an umbrella for methods like supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT) <a class=\"medius-aisummary\" data-starttime=\"00:07:27\">00:07:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:14\">00:08:14</a>. He clarifies that SFT teaches models to imitate provided examples, while RFT rewards models for completing verifiable tasks accurately. Using Foundry’s integration with Ray dashboards, teams can monitor and manage training clusters, visualize job trajectories, and observe fine-tuning rewards in real time <a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:56\">00:10:56</a>. The session concludes by reinforcing that combining SFT and RFT leads to more precise, token-efficient, and production-ready agents. The presenters thank the audience for their patience and summarize the value of data-driven, reinforced model customization <a class=\"medius-aisummary\" data-starttime=\"00:11:21\">00:11:21</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:10 - 21:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:35:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/0ffd6080-1f08-4054-9eea-414215ebfd07","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0ffd6080-1f08-4054-9eea-414215ebfd07","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0ffd6080-1f08-4054-9eea-414215ebfd07","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0ffd6080-1f08-4054-9eea-414215ebfd07","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0ffd6080-1f08-4054-9eea-414215ebfd07","speakerIds":["1768245294609build26-1776895269615001aQWT","1768245294609build26-1776456842448001LPgI"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.853Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0ffd6080-1f08-4054-9eea-414215ebfd07","ogTitle":"Post-Training and Deploying Open Source Reasoning Models in Foundry","ogDescription":"Go beyond prompt engineering to build custom reasoning models using reinforcement learning in Microsoft Foundry. We'll walk through how to train and fine-tune a model to improve reasoning quality, deploy it into Foundry, and integrate it into an agent workflow. Designed for developers comfortable with code, this session focuses on real implementation details, covering training loops, evaluation, and deployment patterns that directly impact agent performance.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776456842448001LPgI","RegistrantKey":"1768245294609build26-1776456842448001LPgI"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776895269615001aQWT","RegistrantKey":"1768245294609build26-1776895269615001aQWT"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"c67ce8b1-fb51-42dc-88ce-39cc3fd1b1c1","startDateTime":"2026-06-02T20:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Freddy Chiu, Vasanth Tovinkere","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774983632180001vRgp","langLocale":"en-US","sessionId":"1774983632180001vRgp","sessionInstanceId":"1774983632180001vRgp","sessionCode":"DEMSP384","sessionCodeNormalized":"EN-US-DEMSP384","title":"Profile and optimize agentic AI on Windows","sortTitle":"profile and optimize agentic ai on windows","description":"Learn how to profile and optimize agentic AI apps on Intel-powered Windows PCs. In this live demo, analyze performance across CPU, GPU, and NPU to identify bottlenecks and improve responsiveness and power efficiency. See how to build performance profiles and apply tuning techniques using Intel Tracing Technology, OpenVINO, and Windows ML.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The session opens with Freddy Chu introducing himself as an AI Frameworks Engineer at Intel, joined by Principal Engineer Vasant Tobinkere <a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>. They outline the talk’s focus on profiling agentic AI workflows on Windows and explain that the session will explore why profiling is important, demonstrate a practical workflow, show profiling of tools, and conclude with a demo and Q&A <a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>. Freddy sets the stage by emphasizing that profiling helps maximize productivity by optimizing prompt iteration speed and efficiency when running large models locally on AI PCs <a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>.</p>\n\n<p><strong>Agentic Workflow Demonstration:</strong> Freddy introduces a sample agentic workflow illustrating how an AI agent designs a 3D wheel model using local computation powered by OpenVINO and the latest Intel Core Ultra architecture <a class=\"medius-aisummary\" data-starttime=\"00:01:24\">00:01:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:12\">00:02:12</a>. The agent reasons through the prompt, generates code, uses OpenSCAD to render the 3D model, and finally stylizes and summarizes the result for continuity in the conversation <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>. Freddy notes this is representative of typical agentic patterns where AI reasons, plans, executes, and remembers interactions. The segue to profiling is introduced by posing the challenge of measuring performance to optimize such multi-step, multimodal workflows <a class=\"medius-aisummary\" data-starttime=\"00:02:55\">00:02:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:06\">00:03:06</a>.</p>\n\n<p><strong>Profiling Methodology and Unified Telemetry:</strong> Vasant begins explaining how to capture and correlate the right telemetry data to make optimization decisions <a class=\"medius-aisummary\" data-starttime=\"00:03:10\">00:03:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:26\">00:03:26</a>. He differentiates between structured, low-overhead platform telemetry (CPU, GPU, NPU, power metrics) and unstructured, higher-overhead application-level telemetry. The talk introduces Intel’s Unified Telemetry framework, which aligns diverse hardware and software data streams into a common time domain for accurate correlation <a class=\"medius-aisummary\" data-starttime=\"00:05:13\">00:05:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>. Vasant discusses its integration with ITT (Instrumentation and Tracing Technology) APIs across Intel’s software stack, enabling developers to observe performance metrics from the hardware driver through the AI middleware to application layers <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>.</p>\n\n<p><strong>Optimization through Hardware Utilization:</strong> Returning to Freddy, the presentation demonstrates profiling results of the 3D modeling workflow, comparing baseline (GPU-only) versus optimized scenarios where workloads are offloaded to the NPU <a class=\"medius-aisummary\" data-starttime=\"00:07:40\">00:07:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>. Offloading stylization and summarization tasks to the NPU allows overlapping computation and response generation, reducing latency and power consumption. The speakers explain how this parallelization improves user experience and extends battery life while maintaining performance efficiency <a class=\"medius-aisummary\" data-starttime=\"00:10:01\">00:10:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>. They further emphasize tuning tool calling and inference profiling as key levers for optimizing increasingly complex agentic workflows.</p>\n\n<p><strong>Tracing, Instrumentation, and Deep Profiling:</strong> Vasant provides deeper insight into how tracing adds structure to unstructured software telemetry, demonstrating how OpenVINO organizes trace data by phases to analyze inference performance <a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:20\">00:14:20</a>. Unified telemetry combines ITT-based software tracing with hardware telemetry into synchronized, time-aligned JSON outputs <a class=\"medius-aisummary\" data-starttime=\"00:14:43\">00:14:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:12\">00:15:12</a>. Developers can instrument their software using ITT APIs to insert lightweight markers that enable transparent data collection once Unified Telemetry is attached <a class=\"medius-aisummary\" data-starttime=\"00:15:44\">00:15:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:17\">00:16:17</a>. These insights make it possible to identify bottlenecks such as model compilation overheads and opportunities to precompile or shift workloads across heterogeneous compute engines.</p>\n\n<p><strong>Demo, Results, and Conclusion:</strong> Freddy’s live demo shows an AI agent profiling Windows ML and ONNX models across multiple device configurations (GPU vs. NPU, FP16 vs. INT8) to generate latency, power, and energy efficiency comparisons <a class=\"medius-aisummary\" data-starttime=\"00:17:25\">00:17:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:55\">00:23:55</a>. The results reveal trade-offs between speed and power consumption, allowing informed workload deployment strategies. Wrapping up <a class=\"medius-aisummary\" data-starttime=\"00:24:26\">00:24:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:24\">00:25:24</a>, the presenters encourage developers to integrate tracing and Unified Telemetry into agentic pipelines to quantify performance, guide optimization, and leverage tools like OpenVINO, ITTAPI, and Windows ML CLI for deeper insights. The talk concludes with an invitation to further discuss Unified Telemetry at Intel’s booth and an evening networking event <a class=\"medius-aisummary\" data-starttime=\"00:25:29\">00:25:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:54\">00:25:54</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"20:00 - 20:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/181d1850-faaf-4f02-b3ce-293fa47f1b5f","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/181d1850-faaf-4f02-b3ce-293fa47f1b5f","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/181d1850-faaf-4f02-b3ce-293fa47f1b5f","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/181d1850-faaf-4f02-b3ce-293fa47f1b5f","speakerIds":["1768245294609build26-1778105929715001G9d4","1768245294609build26-1778105997182001NqLU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.844Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Intel"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/181d1850-faaf-4f02-b3ce-293fa47f1b5f","ogTitle":"Profile and optimize agentic AI on Windows","ogDescription":"Learn how to profile and optimize agentic AI apps on Intel-powered Windows PCs. In this live demo, analyze performance across CPU, GPU, and NPU to identify bottlenecks and improve responsiveness and power efficiency. See how to build performance profiles and apply tuning techniques using Intel Tracing Technology, OpenVINO, and Windows ML.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778105929715001G9d4","RegistrantKey":"1768245294609build26-1778105929715001G9d4"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778105997182001NqLU","RegistrantKey":"1768245294609build26-1778105997182001NqLU"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/DEMSP384_Intel_Profile_Optimize_AgenticAIonWindows_Build26_1780078950713001EptR_1780514949981001XKJl.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"Platform","logicalValue":"Platform"},{"displayValue":"Security in SQL","logicalValue":"Security in SQL"},{"displayValue":"On-prem Migration","logicalValue":"On-prem Migration"},{"displayValue":"Cloud Migration Factory","logicalValue":"Cloud Migration Factory"},{"displayValue":"Analytics","logicalValue":"Analytics"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Linux","logicalValue":"Linux"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"AppTrust","logicalValue":"AppTrust"},{"displayValue":"Platform Security","logicalValue":"Platform Security"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"},{"displayValue":"User Privacy","logicalValue":"User Privacy"},{"displayValue":"Attack Surface Reduction","logicalValue":"Attack Surface Reduction"},{"displayValue":"Cross-device","logicalValue":"Cross-device"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"3010aa68-1838-44d3-ba7d-871c560142ca","startDateTime":"2026-06-03T18:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alexander Wachtel","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290580466001XvvF","langLocale":"en-US","sessionId":"1776290580466001XvvF","sessionInstanceId":"1776290580466001XvvF","sessionCode":"DEM369","sessionCodeNormalized":"EN-US-DEM369","title":"Responsible AI in Action: From Principles to Real‑World Engineering","sortTitle":"responsible ai in action: from principles to real‑world engineering","description":"As AI moves into production, teams must keep models safe, compliant, and reliable without slowing delivery. This session shows how to implement Responsible AI directly in engineering workflows using Microsoft tools. Learn how to assess models, apply safety filters, enforce policies, and monitor risks in real time, plus explore new governance and explainability features that help you build trustworthy AI at scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Purpose of the Demo:</strong>  \nAt <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>, Alex, a Microsoft INDP and trainer, begins by welcoming the audience and introducing a demo focused on Responsible Line Action. He mentions the initiative was successfully run in Germany in January 2026, earning a Microsoft impact award. Alex encourages participants to replicate similar projects within their own Microsoft offices or local communities to enhance discussions around the theme of responsibility. He explains that the session will focus less on redefining AI principles and more on demonstrating practical applications for responsible AI using Microsoft technologies. The goal is to show how responsibility is incorporated through frameworks, standards, and implementation tools available for use across different setups. </p>\n\n<p><strong>Evolution of AI Solutions and Microsoft's Technological Progress:</strong>  \nAround <a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>, Alex transitions into a historical view of Microsoft’s advancements in building AI systems. He recalls early experiments at Microsoft Build events several years ago, where developers were able to build agents to access places like SharePoint Online or Azure Blob Storage. Initially, these solutions worked only as proofs of concept and remained “black box” systems, causing uncertainty about data transparency and control. Over time, Microsoft filled the missing “puzzle pieces” through continuous innovations at Build and Ignite conferences, enabling customers to move safely from proof-of-concept models to production-scale AI solutions. These advancements eliminated the phenomenon of “shadow AI,” turning opaque systems into manageable, measurable, and controllable environments—bringing AI development in alignment with traditional software engineering practices. </p>\n\n<p><strong>Collaboration between Frameworks and Team Members:</strong>  \nBetween <a class=\"medius-aisummary\" data-starttime=\"00:03:53\">00:03:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:03\">00:07:03</a>, Alex introduces his colleagues to illustrate Microsoft’s comprehensive ecosystem. Hannah represents Microsoft Fabric, managing integrated company data and ontologies; Rafael handles Copilot 365 compliance and legal aspects, ensuring adherence to privacy standards. Together, they demonstrate a complete cycle of responsibility from data management to compliance enforcement. Microsoft Fabric plays a key role by providing a unified platform where company data can be modeled as ontologies and queried via natural language agents. These agents can then be quickly published to Copilot in Microsoft 365 with a simple click, extending intelligent capabilities directly to organizational workflows. The updated design of Fabric and Copilot, as noted in Alex’s demo, emphasizes usability and security, making AI integration more intuitive and manageable for any enterprise user. </p>\n\n<p><strong>Compliance, Monitoring, and Purview Integration:</strong>  \nIn the next segment from <a class=\"medius-aisummary\" data-starttime=\"00:09:40\">00:09:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>, Rafael’s participation brings focus to Microsoft Purview, which manages data privacy, compliance, and monitoring across AI services. Purview provides visibility into agent activity and flags sensitive information or unusual interactions. Alex demonstrates how Purview assists in identifying potential misuse or policy violations, such as improper querying of confidential data, by generating logs and compliance alerts. The platform maintains 30-day activity records, enabling organizations to perform advanced hunting through Defender tools to investigate anomalies. The demo showcases Purview’s evolution from a mere compliance system into a sophisticated Data Security Posture Management (DSPM) solution. It highlights how users must monitor and control their tenants, since Microsoft ensures a secure baseline but relies on users to maintain responsible operations of agents and their outputs. </p>\n\n<p><strong>Agent Evaluation, Tracing, and Security Enhancements:</strong>  \nFrom <a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:22:02\">00:22:02</a>, Alex delves into how AI agents are tested, evaluated, and secured inside Microsoft Foundry. He demonstrates tracing and monitoring options—showing metrics like coherence, grounding, token usage, and error rates—all visible through Azure Monitor and Application Insights integration. Developers can perform scheduled evaluations and red teaming tests to ensure security robustness. Using Microsoft’s SDKs, AI models, datasets, and conversational agents can be systematically evaluated and monitored for reliability. Alex shows how automatic evaluations and trace-based data sets simplify iterative improvement. The segment also covers advanced analytics with Kusto Query Language (KQL), allowing users to perform in-depth log hunting similar to SQL but tailored for cloud AI operations. These monitoring features ensure transparency and auditability across all AI tasks, closing the loop on responsible lifecycle management. </p>\n\n<p><strong>Red Teaming, Guardrails, and Conclusion:</strong>  \nIn the final section from <a class=\"medius-aisummary\" data-starttime=\"00:22:03\">00:22:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:38\">00:25:38</a>, the demo illustrates Microsoft’s red teaming process, where agents are tested against harmful prompts and restricted actions. When queries breach defined security policies—such as attempts to retrieve confidential or financial information—the system issues restrictions automatically. Alex demonstrates how developers can reinforce agent safety using custom guardrails and regex-based block lists to prevent specific outputs, like detecting credit card numbers or betting odds. This approach ensures ongoing protection against malicious or unauthorized instructions. He also introduces an AI Gateway for API management that tracks token usage, sets user quotas, and prevents system overloads due to excessive requests—improving cost control and risk mitigation. Alex concludes the demo by emphasizing the importance of responsible management, continuous compliance monitoring, and proactive reinforcement of ethical and secure AI practices within every organization. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:20 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/dc67473a-038c-4b54-a2b6-2219cc574980","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/dc67473a-038c-4b54-a2b6-2219cc574980","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/dc67473a-038c-4b54-a2b6-2219cc574980","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/dc67473a-038c-4b54-a2b6-2219cc574980","speakerIds":["1768245294609build26-1760032715730001qqSb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.904Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/dc67473a-038c-4b54-a2b6-2219cc574980","ogTitle":"Responsible AI in Action: From Principles to Real‑World Engineering","ogDescription":"As AI moves into production, teams must keep models safe, compliant, and reliable without slowing delivery. This session shows how to implement Responsible AI directly in engineering workflows using Microsoft tools. Learn how to assess models, apply safety filters, enforce policies, and monitor risks in real time, plus explore new governance and explainability features that help you build trustworthy AI at scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1760032715730001qqSb","RegistrantKey":"1768245294609build26-1760032715730001qqSb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"29f0f519-4bb3-462d-9f0a-28b21221b92b","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Salil Subbakrishna, Denizhan Yigitbas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709927001rbcd","langLocale":"en-US","sessionId":"1774018709927001rbcd","sessionInstanceId":"1774018709927001rbcd","sessionCode":"DEM301","sessionCodeNormalized":"EN-US-DEM301","title":"Rethinking CI: Actions, AI Agents, and the End of Commit-Fail-Commit","sortTitle":"rethinking ci: actions, ai agents, and the end of commit-fail-commit","description":"Your pipelines automate the predictable. But what about triaging issues, reviewing PRs, responding to incidents, and coordinating across tools? See what's new in GitHub Actions and how it's becoming the execution layer for AI agents across your dev lifecycle. We'll cover agent-triggered workflows, MCP server integration, and automated handoffs that keep humans in the loop — plus how to finally break the 'commit - see CI fail - commit again' loop.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with Salil and Dennis, both product managers at GitHub, introducing themselves and the theme of the talk: rethinking continuous integration (CI) through GitHub Actions, AI, and agentic workflows <a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>. Salil highlights how GitHub Actions usage has exploded from 2.8 million to 850 million jobs per week within a short timespan, due to the adoption of AI and automation <a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>. Despite this growth, he acknowledges a recurring user frustration called the \"commit-fail-commit\" cycle, where developers must repeatedly push changes to fix failed workflows. GitHub aims to break this pattern by introducing two major tools: agent-based workflows and a new Actions Debugger for direct workflow troubleshooting <a class=\"medius-aisummary\" data-starttime=\"00:01:47\">00:01:47</a>.</p>\n\n<p><strong>Agentic Workflows and the “CI Doctor” Concept:</strong> Salil explains agentic workflows — AI-infused automations written in Markdown that GitHub compiles into standard Actions YAML files <a class=\"medius-aisummary\" data-starttime=\"00:02:13\">00:02:13</a>. The demonstration centers on a special agentic workflow called \"CI Doctor,\" designed to automatically analyze failed runs and identify root causes. After showing a typical Python-based CI failure, Salil triggers the CI Doctor workflow, which detects the failure and automatically launches an agent that produces diagnostic information and actionable recommendations <a class=\"medius-aisummary\" data-starttime=\"00:03:23\">00:03:23</a>. The CI Doctor inspects failure patterns, assesses logs, and creates an issue summary with reasons for the failure and suggested code fixes, such as converting string inputs to integers and adding validation steps <a class=\"medius-aisummary\" data-starttime=\"00:08:08\">00:08:08</a>. This automation significantly reduces developer intervention and shortens the debugging cycle.</p>\n\n<p><strong>Creating and Customizing Agentic Workflows:</strong> Salil dives deeper into how these workflows can be defined, compiled, and extended. Developers create `.md` files that describe AI logic, compile them using the GitHub CLI extension command `github aw compile`, and produce a `.lock.yaml` file that GitHub Actions executes <a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a>. He demonstrates live customization: using GitHub Copilot to modify prompts, add slash commands, and tailor workflows for specific use cases <a class=\"medius-aisummary\" data-starttime=\"00:05:35\">00:05:35</a>. These workflows can be versioned, shared across repositories, and refined collaboratively, enabling teams to build their own diagnostic or triage agents. GitHub also provides prebuilt templates—for CI analysis, issue triage, and documentation automation—to accelerate adoption and reuse <a class=\"medius-aisummary\" data-starttime=\"00:07:45\">00:07:45</a>.</p>\n\n<p><strong>Introducing the Actions Debugger:</strong> After showing how agentic workflows handle automated analysis, Salil transitions to the upcoming Actions Debugger, designed for deeper human-in-the-loop investigation when workflows behave unpredictably <a class=\"medius-aisummary\" data-starttime=\"00:10:49\">00:10:49</a>. Using a live PR example, he demonstrates connecting the debugger from VS Code to an active GitHub Actions runner through the Debug Adapter Protocol (DAP) <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. Once connected, the debugger exposes all runtime variables, environment data, and logs in real time. Salil traces an issue where only one file was processed in a PR; by inspecting the `base_sha` and `head_sha` variables directly on the runner, he identifies a logic error in the diff comparison and confirms the fix instantly <a class=\"medius-aisummary\" data-starttime=\"00:17:07\">00:17:07</a>. This hands-on approach allows developers to explore without repeated commits, turning Actions debugging into an interactive, IDE-level experience.</p>\n\n<p><strong>Demonstration Outcome and Debugging Advantages:</strong> The debugger session concludes successfully, with Salil pinpointing the root cause of the workflow issue and verifying the corrected logic <a class=\"medius-aisummary\" data-starttime=\"00:19:29\">00:19:29</a>. He shows how the DAP-based integration supports standard debugging operations—running commands, viewing contexts, and examining variables securely across various clients such as VS Code or Neovim <a class=\"medius-aisummary\" data-starttime=\"00:16:35\">00:16:35</a>. This approach eliminates the need for “print statement” debugging and allows the same step-by-step tracing familiar in traditional software debugging, now applied directly to CI pipelines. Dennis encapsulates this as a true “debugger for runners,” reinforcing how these innovations transform CI troubleshooting from guesswork into an interactive, efficient process <a class=\"medius-aisummary\" data-starttime=\"00:20:05\">00:20:05</a>.</p>\n\n<p><strong>Roadmap, Availability, and Closing Remarks:</strong> The presentation wraps up with announcements on release timelines <a class=\"medius-aisummary\" data-starttime=\"00:20:40\">00:20:40</a>. Agentic workflows will enter public preview the following week, supported by documentation and quick-start examples for issue management, documentation automation, and multi-repository orchestration. The Actions Debugger, meanwhile, remains an internal project while GitHub focuses on maintaining strong security measures before public release <a class=\"medius-aisummary\" data-starttime=\"00:21:20\">00:21:20</a>. The speakers encourage the audience to provide feedback through GitHub’s public forums to influence future templates and features. They close by thanking participants and emphasizing GitHub’s commitment to improving developer experience through AI-driven automation and transparent tooling innovations <a class=\"medius-aisummary\" data-starttime=\"00:22:19\">00:22:19</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:00 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/079bf23f-302e-4b23-bc15-add6bb85deee","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/079bf23f-302e-4b23-bc15-add6bb85deee","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/079bf23f-302e-4b23-bc15-add6bb85deee","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/079bf23f-302e-4b23-bc15-add6bb85deee","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/079bf23f-302e-4b23-bc15-add6bb85deee","speakerIds":["1768245294609build26-1774880330948001FAkq","1768245294609build26-1775278306002001tw5Z"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.789Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/079bf23f-302e-4b23-bc15-add6bb85deee","ogTitle":"Rethinking CI: Actions, AI Agents, and the End of Commit-Fail-Commit","ogDescription":"Your pipelines automate the predictable. But what about triaging issues, reviewing PRs, responding to incidents, and coordinating across tools? See what's new in GitHub Actions and how it's becoming the execution layer for AI agents across your dev lifecycle. We'll cover agent-triggered workflows, MCP server integration, and automated handoffs that keep humans in the loop — plus how to finally break the 'commit - see CI fail - commit again' loop.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774880330948001FAkq","RegistrantKey":"1768245294609build26-1774880330948001FAkq"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775278306002001tw5Z","RegistrantKey":"1768245294609build26-1775278306002001tw5Z"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"OSS CI/CD Libraries","logicalValue":"OSS CI/CD Libraries"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"4df2d5b4-b077-4bfe-af66-b8a6c1e9f870","startDateTime":"2026-06-02T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gova Babu, Pranay Bakre, Sameer Nori","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774391978124001tcWQ","langLocale":"en-US","sessionId":"1774391978124001tcWQ","sessionInstanceId":"1774391978124001tcWQ","sessionCode":"DEMSP381","sessionCodeNormalized":"EN-US-DEMSP381","title":"Scale agentic AI cost‑efficiently on Azure with Arm Cobalt VMs","sortTitle":"scale agentic ai cost‑efficiently on azure with arm cobalt vms","description":"As applications evolve into agent-driven systems, inference must scale efficiently. In this session, Arm and Microsoft show you how the latest Azure Cobalt VMs enable cost-effective, CPU-based AI for agentic and cloud-native workloads. In a live AKS demo, you'll learn how to use Azure Cobalt VMs to deploy LLM inferencing and app tiers, with insights on performance, scaling, and real-world design patterns.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Introduction and Overview:</strong> The session begins with greetings and acknowledgments for the audience’s patience after some technical difficulties (00:00:00–00:00:04). Sameer Nori from Arm introduces himself and outlines the agenda, which includes discussions on partnerships, the Arm software ecosystem, and a demo highlighting how identity workloads can be scaled cost-effectively on Azure Virtual Machines (00:00:07–00:00:31). He explains Arm’s long history in computing and notes their collaboration with Microsoft and Azure, emphasizing their role in powering devices from cloud to edge (00:00:35–00:00:58).</p>\n\n<p><strong>Arm–Microsoft Partnership and Technology Foundations:</strong> Sameer elaborates on the partnership’s two pillars: silicon innovation and software enablement (00:01:01–00:01:18). Under silicon innovation, Arm collaborates with Microsoft hardware teams to develop the Cobalt chip series offering notable improvements in price-performance ratios—Cobalt 200 delivers 50% better performance than Cobalt 100 (00:01:23–00:01:35). On software enablement, there are more than 22 million developers building on Arm, with nearly all CNCF projects supporting Arm architectures, spanning open source, ISV packages, Linux distributions, and cloud-native frameworks such as AIML and SaaS platforms (00:01:39–00:02:06).</p>\n\n<p><strong>Cobalt 200 Virtual Machines – Performance and Capabilities:</strong> Gova from Microsoft’s Arm product team takes over to discuss the evolution from Cobalt 100 to Cobalt 200 Virtual Machines (00:02:26–00:02:33). He highlights customer adoption successes for Cobalt 100, including Microsoft Teams and Defender, which use Arm-based Azure VMs for robust workloads (00:03:04–00:03:17). Gova announces that Cobalt 200 VMs, built on 3nm Arm architecture and optimized for Microsoft workloads, now ship with Azure Boost and achieve around 50% better per-core performance compared to the previous generation (00:03:35–00:04:12). These VMs support multiple configurations—compute, memory-optimized, and dense storage variants—to cover both enterprise and cloud-native demands (00:04:27–00:05:01).</p>\n\n<p><strong>Benchmarking and AI Readiness:</strong> Gova shares benchmark results demonstrating that Cobalt 200 consistently exceeds previous generations across industry and Microsoft-specific metrics (00:05:07–00:05:39). He emphasizes that Microsoft’s own products validate these results through direct use of Arm-powered Azure VMs. Gova then shifts focus toward the future—agentic AI workloads—asserting that Cobalt 200 is engineered to efficiently handle AI-driven processes like sandbox creation and complex request loops without compromising cost-performance efficiency (00:06:00–00:06:25). The VMs are now available in preview across eight regions, with more expansion planned upon general availability (00:06:33–00:06:46).</p>\n\n<p><strong>Demo: Cloud-Native and AI Agent Workflows:</strong> Pranay leads a demonstration showcasing how cloud-native applications can operate seamlessly on Cobalt-powered clusters (00:07:05–00:07:21). He explains how traditional workflows evolve into AI-first and agentic workflows enabled by both Cobalt 100 and 200 VMs, supporting distributed microservices architectures and inference workloads within secure clusters (00:07:58–00:09:02). The demo features a polyglot shopping cart microservice that uses CPU-based inferencing on Cobalt 200 VMs, demonstrating local data processing without reliance on external AI models (00:09:11–00:11:22).</p>\n\n<p><strong>Conclusion and Engagement Opportunities:</strong> As the demo wraps up, Pranay displays how an AI-driven shopping interface maintains context through local inference, exemplifying secure and efficient CPU-based computation (00:13:40–00:15:09). He highlights upcoming hands-on lab sessions and invites participants to explore Arm Cloud Migration resources, which assist in transitioning workloads to Arm64 with performance analysis and engineering support (00:16:02–00:16:38). The session concludes with appreciation for attendees and encouragement to engage further with the Arm and Microsoft teams for migration and optimization assistance (00:16:39).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:30 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/2796f78a-ca86-49c1-98cc-ba57308ea186","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/2796f78a-ca86-49c1-98cc-ba57308ea186","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/2796f78a-ca86-49c1-98cc-ba57308ea186","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/2796f78a-ca86-49c1-98cc-ba57308ea186","speakerIds":["1768245294609build26-1776877272377001WqB6","1768245294609build26-1776197975793001dvpl","1768245294609build26-1776814583002001KGlJ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.636Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Arm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/2796f78a-ca86-49c1-98cc-ba57308ea186","ogTitle":"Scale agentic AI cost‑efficiently on Azure with Arm Cobalt VMs","ogDescription":"As applications evolve into agent-driven systems, inference must scale efficiently. In this session, Arm and Microsoft show you how the latest Azure Cobalt VMs enable cost-effective, CPU-based AI for agentic and cloud-native workloads. In a live AKS demo, you'll learn how to use Azure Cobalt VMs to deploy LLM inferencing and app tiers, with insights on performance, scaling, and real-world design patterns.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776197975793001dvpl","RegistrantKey":"1768245294609build26-1776197975793001dvpl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776877272377001WqB6","RegistrantKey":"1768245294609build26-1776877272377001WqB6"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776814583002001KGlJ","RegistrantKey":"1768245294609build26-1776814583002001KGlJ"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/DEMSP381%20-%20Arm%20-%20Microsoft%20Build%202026%20Theatre%20session_1780097441922001IBRE_1780512629516001chce.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"AKS","logicalValue":"AKS"}]},{"@search.score":1.0,"documentId":"6dc4d1a7-c1db-44f8-bef2-407dec69bd6e","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Poorvi Narang, Jim Perrin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018705970001rNQ0","langLocale":"en-US","sessionId":"1774018705970001rNQ0","sessionInstanceId":"1774018705970001rNQ0","sessionCode":"DEM311","sessionCodeNormalized":"EN-US-DEM311","title":"Scale cloud-native workloads with Azure Linux","sortTitle":"scale cloud-native workloads with azure linux","description":"Azure Linux is a purpose-built Linux distribution optimized for Azure. Learn how Azure Linux supports cloud-native and AI workloads with deep integration into the Azure ecosystem, delivering a consistent Linux platform across containers and other Azure compute services. Designed with a minimal footprint, Azure Linux enables faster provisioning and scaling, reduces the attack surface, and incorporates Azure's robust cloud security standards.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The presentation opens with Jim Perrin and Poorvinder Rang introducing themselves as program managers on the Azure Linux team (<a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>). They briefly outline the session: an overview of Azure Linux, insights into Azure Container Linux for Kubernetes and container users, a product demo, and time for questions. The goal, as Perrin explains, is to help attendees understand what Azure Linux is, what Microsoft is building, and how it integrates with both containerized and virtualized environments within Azure (<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>). The introduction sets the stage for a technical deep dive into the evolution of Azure’s Linux distribution strategy.</p>\n\n<p><strong>Azure Linux Foundations and Fedora Partnership:</strong> Early in the session, Jim details how Azure Linux derives from Fedora, echoing the upstream–downstream relationship that Red Hat Enterprise Linux maintains (<a class=\"medius-aisummary\" data-starttime=\"00:00:50\">00:00:50</a>). Microsoft partners closely with the Fedora community, contributing upstream while maintaining transparency in how Azure Linux deviates from Fedora via public GitHub repositories (<a class=\"medius-aisummary\" data-starttime=\"00:01:42\">00:01:42</a>). The use of Fedora as an upstream source streamlines ecosystem alignment and brings existing third-party support. Jim emphasizes Microsoft’s added value around compliance—like FIPS and FedRAMP certifications—and the concept of a consistent cadence for updates and vulnerability management modeled after the well-known “Patch Tuesday” concept (<a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>). This consistency ensures reliability across workloads and security levels for enterprises adopting Azure Linux.</p>\n\n<p><strong>Architecture and Kernel Optimization for Azure:</strong> Jim transitions into the technical structure defining both Azure Linux and Azure Container Linux (<a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>). Both share identical binaries for compatibility, tuned specifically for Azure performance, reliability, and hardware enablement. The kernel supports both long-term stability and rapid iteration cycles as required by different customers (<a class=\"medius-aisummary\" data-starttime=\"00:03:33\">00:03:33</a>). These optimizations combine open-source flexibility with Azure’s managed fleet consistency, integrating natively with tools such as Azure Update Manager. This architectural coherence ensures customers can deploy workloads seamlessly across virtual machines, containers, or immutable infrastructure images without revalidation, preserving environment parity across the ecosystem.</p>\n\n<p><strong>Live Demonstration Across Environments:</strong> Poorvi leads the live demo beginning with Azure Linux 4.0 running on Windows Subsystem for Linux (WSL) (<a class=\"medius-aisummary\" data-starttime=\"00:05:21\">00:05:21</a>). She runs a simple Python application to demonstrate that the same app behaves identically in WSL, a virtual machine, and a Kubernetes deployment via Azure Kubernetes Service (AKS). During the VM portion, she highlights integrated security features such as SELinux enforced by default, lean image sizes designed through “declarative deviations,” and removal of nonessential desktop packages for hardened, cloud-native security (<a class=\"medius-aisummary\" data-starttime=\"00:09:39\">00:09:39</a>). The demo also spotlights the new DNF 5 package manager for faster dependency handling, default firewall rules with explicit manual control, and preconfigured security postures. This portion underlines the central Azure Linux tenet: strong out-of-the-box security and consistency from developer workstation to cloud infrastructure (<a class=\"medius-aisummary\" data-starttime=\"00:13:36\">00:13:36</a>).</p>\n\n<p><strong>Azure Container Linux and Immutable Infrastructure:</strong> Jim then elaborates on Azure Container Linux as an immutable derivative built from Flatcar Container Linux, contributed by Microsoft to the CNCF (<a class=\"medius-aisummary\" data-starttime=\"00:18:40\">00:18:40</a>). He clarifies that Flatcar is not being deprecated—Azure Container Linux simply provides an enterprise-ready, productized variant aligned with Azure’s operational model. The distribution maintains consistency with Azure Linux, featuring secure boot, dm-verity validation, read-only file systems, and SELinux enforcement. The live deployment demo on AKS showcases automatic verification, minimal kernel variations, and secure configuration defaults (<a class=\"medius-aisummary\" data-starttime=\"00:21:35\">00:21:35</a>). Azure Container Linux 3.0 is now generally available, with version 4.0 entering preview alignment. Together, these editions unify the immutable infrastructure experience within Azure’s container ecosystem.</p>\n\n<p><strong>Q&A and Closing Remarks:</strong> In closing (<a class=\"medius-aisummary\" data-starttime=\"00:22:32\">00:22:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:27\">00:25:27</a>), Jim answers an audience query regarding firewall D rule management, assuring clear documentation of any package deviations in GitHub. The presenters summarize product availability: Azure Linux 4.0 in public preview, with Azure Container Linux officially GA. They invite collaboration, feedback, and partnerships through GitHub and community calls. The session concludes with encouragement for developers and partners to test, validate, and integrate Azure Linux across WSL, VM, and Kubernetes deployments, emphasizing Microsoft’s commitment to transparent open-source stewardship and secure, universal Linux experience in the Azure cloud.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/1b886f16-d29b-4873-a32c-b8002baa7b46","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/1b886f16-d29b-4873-a32c-b8002baa7b46","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/1b886f16-d29b-4873-a32c-b8002baa7b46","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/1b886f16-d29b-4873-a32c-b8002baa7b46","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/1b886f16-d29b-4873-a32c-b8002baa7b46","speakerIds":["1768245294609build26-1775680212026001g2dP","1768245294609build26-1774284997117001po3o"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.461Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/1b886f16-d29b-4873-a32c-b8002baa7b46","ogTitle":"Scale cloud-native workloads with Azure Linux","ogDescription":"Azure Linux is a purpose-built Linux distribution optimized for Azure. Learn how Azure Linux supports cloud-native and AI workloads with deep integration into the Azure ecosystem, delivering a consistent Linux platform across containers and other Azure compute services. Designed with a minimal footprint, Azure Linux enables faster provisioning and scaling, reduces the attack surface, and incorporates Azure's robust cloud security standards.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C","logicalValue":"C"}],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284997117001po3o","RegistrantKey":"1768245294609build26-1774284997117001po3o"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775680212026001g2dP","RegistrantKey":"1768245294609build26-1775680212026001g2dP"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Linux","logicalValue":"Azure Linux"}]},{"@search.score":1.0,"documentId":"cb4b6744-e0b4-469d-a87c-eab03432497c","startDateTime":"2026-06-02T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sam Basu, Colin Whitlatch","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776443414500001FKs4","langLocale":"en-US","sessionId":"1776443414500001FKs4","sessionInstanceId":"1776443414500001FKs4","sessionCode":"DEMSP394","sessionCodeNormalized":"EN-US-DEMSP394","title":"Scale enterprise .NET apps with AI‑assisted cross‑platform workflows","sortTitle":"scale enterprise .net apps with ai‑assisted cross‑platform workflows","description":"Most AI tooling for .NET targets greenfield code generation. Enterprise app development is different: layered systems, cross-platform targets, mixed-skill teams, and production demands. In this session, Uno Platform and Kahua show an end-to-end AI-assisted workflow across web, desktop, and mobile, where AI can inspect live app state, understand runtime UI behavior, and act with context. Be ready for real productivity with contextual AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The video opens with greetings at Microsoft Build (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>), where Sam Basu from Uno Platform and Colin Whitlatch from Kahua introduce their joint session on \"Cross-platform modern .NET with Uno Platform at enterprise scale with AI-assisted workflows.\" They explain their collaboration—Uno as the platform maker and Kahua as an enterprise-scale user pushing its capabilities. The speakers position the talk as an exploration of how developer tooling is evolving to support agentic, AI-enhanced workflows in real-world applications, setting the tone for multiple announcements during the session.</p>\n\n<p><strong>AI-Driven Development Evolution:</strong> Early discussion (<a class=\"medius-aisummary\" data-starttime=\"00:00:50\">00:00:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>) centers on how software building has progressed from manual coding to visual tools and now to AI-assisted approaches. The presenters outline a spectrum from code-centric practices to high-level automation, emphasizing that AI integration makes development faster but that solid fundamentals like architecture and context remain critical. Kahua’s experience exemplifies this transition—they’ve moved from traditional development to fully AI-orchestrated coding, where developers serve more as orchestrators than line-by-line coders. The team highlights integrating AI with design tools like Figma and extending support for end-to-end workflows that produce runnable enterprise code, not prototypes.</p>\n\n<p><strong>Introducing Uno Platform Studio and Tooling:</strong> The speakers then present Uno Platform Studio as a suite of AI and design tools for modern .NET development (<a class=\"medius-aisummary\" data-starttime=\"00:03:16\">00:03:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:55\">00:04:55</a>). They explain features such as “Hot Design” for real-time UI tweaking, seamless integration with Figma pipelines, and AI agents with full application context through MCP tools. The goal is to let developers and AI co-edit live applications that run across iOS, Android, WebAssembly, Windows, Linux, and macOS from one shared code base. This segment emphasizes context-driven AI, ensuring that generated code aligns with project documentation and user interaction patterns—what they call “AI with eyes and hands.”</p>\n\n<p><strong>Uno Platform Studio 3.0 Announced and Demoed:</strong> A major announcement follows—the launch of Uno Platform Studio 3.0 (<a class=\"medius-aisummary\" data-starttime=\"00:05:08\">00:05:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>). The update introduces a specialized agent equipped with more than 60 skills, AI plugins, and deep MCP integration, enabling intelligent guidance in app creation. The session transitions into a live demo where AI builds a weather app directly in the browser using a single text prompt. Viewers see how apps can shift between light and dark modes, preview on various devices, and be exported into Visual Studio or GitHub seamlessly. The platform demonstrates full responsiveness, real-time collaboration through hot reload across multiple devices, and easy design iteration. Subsequent examples include CRM and coffee-ordering applications rendered live, underscoring Studio 3.0’s full-stack generation capability.</p>\n\n<p><strong>Advanced Use Cases and AI Integration:</strong> The demo extends into advanced scenarios (<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>), showing Uno integration within Visual Studio Code and GitHub Copilot. The agent autonomously launches, edits, and validates applications through visual trees and interactive design modes. Demonstrations show live UI testing, data-driven chart updates, and AI performing validation steps—mirroring concepts from end-to-end UI testing frameworks. Colin emphasizes customizable AI orchestration and cost-efficient model use, referring to GPT-5 Mini as an example. Core developer utilities such as hot reload, hot design, and doc-grounded MCP servers are shown to improve productivity and iteration, all while preserving native .NET standards like XAML and C#.</p>\n\n<p><strong>Future Vision and Conclusion:</strong> The final section (<a class=\"medius-aisummary\" data-starttime=\"00:17:40\">00:17:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:33\">00:23:33</a>) summarizes new building blocks in Uno Platform Studio 3.0—skills, plugins, design previews, and snippet-based UI assembly. Kahua shares real-world enterprise implementations using AI-driven workflows for large-scale construction and inspection apps leveraging LiDAR and 360-degree visualization. Both teams urge developers to adopt a human-and-AI collaborative model: humans define goals while AI iterates and verifies output in continuous loops. They close by stressing that while AI accelerates development, foundational principles and context remain cornerstones of reliable software. Attendees are invited to explore cross-platform demos and meet the teams at their Microsoft Build booth for interactive sessions and even themed arcade games built with Uno Platform technology.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:40 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/90f6fe71-8cba-465f-b53c-a3ed30e154c6","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/90f6fe71-8cba-465f-b53c-a3ed30e154c6","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/90f6fe71-8cba-465f-b53c-a3ed30e154c6","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/90f6fe71-8cba-465f-b53c-a3ed30e154c6","speakerIds":["1768245294609build26-1776201109444001xetr","1768245294609build26-1776884774053001DbiF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.382Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Uno Platform"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/90f6fe71-8cba-465f-b53c-a3ed30e154c6","ogTitle":"Scale enterprise .NET apps with AI‑assisted cross‑platform workflows","ogDescription":"Most AI tooling for .NET targets greenfield code generation. Enterprise app development is different: layered systems, cross-platform targets, mixed-skill teams, and production demands. In this session, Uno Platform and Kahua show an end-to-end AI-assisted workflow across web, desktop, and mobile, where AI can inspect live app state, understand runtime UI behavior, and act with context. Be ready for real productivity with contextual AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776201109444001xetr","RegistrantKey":"1768245294609build26-1776201109444001xetr"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776884774053001DbiF","RegistrantKey":"1768245294609build26-1776884774053001DbiF"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/BuildSession_1780197239629001y8vE_1780513210138001nIOd.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"}]},{"@search.score":1.0,"documentId":"594b56c6-efa6-4ebe-b292-2649dec58a64","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ali Golshan, Alex Watson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775842629438001AJOw","langLocale":"en-US","sessionId":"1775842629438001AJOw","sessionInstanceId":"1775842629438001AJOw","sessionCode":"DEMSP387","sessionCodeNormalized":"EN-US-DEMSP387","title":"Secure agent workflows in GitHub Copilot with NVIDIA OpenShell","sortTitle":"secure agent workflows in github copilot with nvidia openshell","description":"Agentic workflows in GitHub Copilot can run complex tasks with broad access to local systems. In this session, see how to enforce policy and governance for agents operating in CLI environments. You’ll walk through how NVIDIA OpenShell applies kernel-level controls, isolates agent processes, and limits permissions based on task context. Learn how to apply zero-trust principles to secure agent-driven development workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Project Overview:</strong> The video opens with Ali and Alex introducing <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a> Open Shell, a project developed at NVIDIA. They explain that the focus is on building an agent-native software stack and exploring what software should look like in an environment where agents execute long-running, stateful sessions. The main goals for Open Shell are developing a secure, trusted runtime and creating an environment designed for machine-speed operation. The project was initiated to handle autonomy and persistence for large-scale agent deployments while maintaining zero trust and strong governance.</p>\n\n<p><strong>Security Foundations and Architecture:</strong> The discussion moves to security principles, where the team highlights the zero-trust environment of Open Shell <a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:12\">00:02:12</a>. Each sandbox starts locked down and only gains access as explicitly granted. Policies and governance are enforced at the kernel or infrastructure level rather than the model layer. Examples, like preventing prompt injection, show that credentials and secrets reside outside the sandbox to limit exposure. The system architecture is fully open source under Apache 2.0 <a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>, with gateways, sandboxes, and sub-agent isolation forming the core primitives that ensure security and scalability.</p>\n\n<p><strong>Policy Prover and Privacy Router:</strong> Alex introduces the Policy Prover component <a class=\"medius-aisummary\" data-starttime=\"00:04:24\">00:04:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:54\">00:05:54</a>, which uses formal logic verification based on OPA and Rego policy languages. This system can mathematically prove an agent’s permissions and prevent vulnerabilities caused by human error or false approvals. The privacy router <a class=\"medius-aisummary\" data-starttime=\"00:07:41\">00:07:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:23\">00:09:23</a> is another unique component that uses technology derived from synthetic data protection. It automatically routes queries to appropriate inference models depending on whether personally identifiable information (PII) is detected, allowing secure orchestration among local and frontier models. This balance enables efficient use of global compute resources while keeping sensitive data local.</p>\n\n<p><strong>Demo: GitHub Persistence and Agent Collaboration:</strong> The first demo <a class=\"medius-aisummary\" data-starttime=\"00:09:42\">00:09:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:06\">00:15:06</a> shows how multiple agents can use GitHub as a persistence layer within Open Shell. Several sandboxes are spun up to execute tasks independently, each storing results in isolated GitHub files. A synthesis agent then aggregates these outcomes. During the demonstration, the team highlights policy management for GitHub access, ensuring that agents only interact through predefined REST endpoints without handling credentials directly. Hot reloading of policies is emphasized as a design feature optimizing agent-speed execution, preventing sandbox restarts and latency issues.</p>\n\n<p><strong>Demo: Automated Policy Negotiation:</strong> In the second demo <a class=\"medius-aisummary\" data-starttime=\"00:15:35\">00:15:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>, Ali and Alex demonstrate dynamic policy negotiation. An agent with only read access attempts to write to GitHub, prompting automatic blocking and self-proposal for additional permissions. The Policy Prover evaluates the request, proofs the logic against existing rules, and either approves or denies the expansion. All actions are logged in a standardized cybersecurity format (OCSF), compatible with tools like Splunk and Datadog. This automation drastically reduces human intervention while maintaining granular control over agent privileges.</p>\n\n<p><strong>Conclusion and Future Outlook:</strong> The presentation closes <a class=\"medius-aisummary\" data-starttime=\"00:18:11\">00:18:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:45\">00:21:45</a> with details on Open Shell’s roadmap and partnerships. The project is already integrated into major platforms including Azure, GitHub, Ubuntu, and Red Hat OpenShift. It remains open source under Apache 2.0, with plans for CNCF or Linux Foundation contribution. The speakers invite community testing and contributions, noting that the alpha version debuted at GTC in March and will advance to beta soon. They explain new driver capabilities for sandboxing primitives such as Firecracker or Kubernetes SIG sandboxing. The session ends with questions on reducing human approval rates and maintaining secure boundaries through dynamic policy negotiation.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"16:30 - 16:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b73f26f9-8342-4119-b417-82d0136435b9","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b73f26f9-8342-4119-b417-82d0136435b9","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b73f26f9-8342-4119-b417-82d0136435b9","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b73f26f9-8342-4119-b417-82d0136435b9","speakerIds":["1768245294609build26-1778198381538001QtyE","1768245294609build26-1780429101773001j9MJ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.901Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["NVIDIA"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b73f26f9-8342-4119-b417-82d0136435b9","ogTitle":"Secure agent workflows in GitHub Copilot with NVIDIA OpenShell","ogDescription":"Agentic workflows in GitHub Copilot can run complex tasks with broad access to local systems. In this session, see how to enforce policy and governance for agents operating in CLI environments. You’ll walk through how NVIDIA OpenShell applies kernel-level controls, isolates agent processes, and limits permissions based on task context. Learn how to apply zero-trust principles to secure agent-driven development workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778198381538001QtyE","RegistrantKey":"1768245294609build26-1778198381538001QtyE"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780429101773001j9MJ","RegistrantKey":"1768245294609build26-1780429101773001j9MJ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"}]},{"@search.score":1.0,"documentId":"6bdebefd-52d2-4bdf-ad5a-e5c0d68d7141","startDateTime":"2026-06-02T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Marko Hotti, Sergiy Smyrnov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018705815001rKG3","langLocale":"en-US","sessionId":"1774018705815001rKG3","sessionInstanceId":"1774018705815001rKG3","sessionCode":"DEM310","sessionCodeNormalized":"EN-US-DEM310","title":"Ship code faster with AI-powered NoSQL schema design","sortTitle":"ship code faster with ai-powered nosql schema design","description":"NoSQL schema design is hard—denormalization decisions, partition key selection, and data modeling patterns require expertise. Use GitHub Copilot and the new Azure Cosmos DB Agent Toolkit to accelerate development with AI-assisted schema generation, query optimization suggestions, and refactoring recommendations. Iterate rapidly with the new Mac/Linux emulator for local testing. Demo shows schema evolution across three iterations in 30 minutes versus days of manual design.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Session Opening and Introduction:</strong> The video begins with greetings and welcomes from the hosts <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>. Marco and Sergei are introduced as the speakers who will discuss how to ship code faster using AI-powered NoSQL schema design. Marco starts the session by engaging the audience and asking how many are developing NoSQL applications with Azure Cosmos DB or other technologies such as MongoDB and DynamoDB <a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:36\">00:00:36</a>. He highlights the appeal of NoSQL databases for agile schema iteration using JSON and explains how Azure Cosmos DB is optimized for low-latency, scalable applications, suitable for both operational data and AI-native workloads <a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:11\">00:01:11</a>.</p>\n\n<p><strong>Overview of Azure Cosmos DB and AI Integration:</strong> Marco expands on the capabilities of Azure Cosmos DB, mentioning integrated vector search, full-text search, hybrid semantic ranking, and strong integration with Microsoft Foundry for hosting large language models <a class=\"medius-aisummary\" data-starttime=\"00:01:23\">00:01:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:46\">00:01:46</a>. He emphasizes the platform’s suitability for globally distributed AI workloads and provides real-world examples like OpenAI’s use of Cosmos DB for ChatGPT and ServiceNow deployments <a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>. Marco transitions to how developers can use tools such as GitHub Copilot and MCP integration to accelerate application building with AI assistance, underscoring the importance of a standardized, secure connection method for apps and agents <a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:39\">00:02:39</a>. He concludes the overview and hands the session over to Sergei for a demo showcasing Azure Cosmos DB Agent Kit <a class=\"medius-aisummary\" data-starttime=\"00:02:42\">00:02:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:54\">00:02:54</a>.</p>\n\n<p><strong>Demo Setup and Conceptual Foundations:</strong> Sergei introduces the demo, focused on a classic e-commerce model including customer, product, and sales order domains <a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>. After regaining connection mid-demo, he explains that traditional modernization often converts relational tables directly into containers, which while valid, is suboptimal. His goal is to demonstrate how the Cosmos DB Agent Kit can improve schema efficiency using AI-guided design <a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:41\">00:04:41</a>. The Agent Kit includes best practices and Copilot skills that can be installed locally through simple commands and used across different environments <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:11\">00:05:11</a>. He stresses how indexing and skill-based optimization empower the agent to evaluate schemas for NoSQL systems based on various SDKs.</p>\n\n<p><strong>Defining Access Patterns and Optimization Strategy:</strong> Sergei demonstrates how adding domains and defining structured access patterns allows the AI agent to design for both model and real-world query efficiency <a class=\"medius-aisummary\" data-starttime=\"00:06:35\">00:06:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>. Four main access patterns are defined: retrieving customers with recent orders, fetching single order details, performing CRUD operations on orders, and listing products by category <a class=\"medius-aisummary\" data-starttime=\"00:07:04\">00:07:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:23\">00:07:23</a>. He discusses volumetric projections to anticipate scale from 10 development users to millions in production, showing how early design decisions affect cost and performance <a class=\"medius-aisummary\" data-starttime=\"00:07:42\">00:07:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>. The demo compares naive schema design, where each table becomes an individual container, with optimized approaches that merge domains and apply consistent partition strategies <a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:50\">00:09:50</a>.</p>\n\n<p><strong>Demonstration of Agent-Guided Schema and Data Validation:</strong> Using the Agent Kit, Sergei runs AI-guided iterations that propose collapsing related domains (customers with orders, products with categories) to improve partitioning and indexing <a class=\"medius-aisummary\" data-starttime=\"00:14:46\">00:14:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:19\">00:15:19</a>. He then deploys this optimized schema locally using the Azure Cosmos DB Emulator, iterating on container structures and verifying configurations <a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:20\">00:16:20</a>. The emulator’s fast iteration allows validation of indexing policies and document shapes through CSV-driven data conversion and JSON validation routines <a class=\"medius-aisummary\" data-starttime=\"00:18:05\">00:18:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:22\">00:18:22</a>. Sergei shows how data types are standardized and validated, resulting in clean combined customer and order documents optimized for query efficiency <a class=\"medius-aisummary\" data-starttime=\"00:19:53\">00:19:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:22\">00:20:22</a>.</p>\n\n<p><strong>Results, Cost Savings, and Conclusion:</strong> In closing, Sergei compares performance metrics between naive and optimized iterations <a class=\"medius-aisummary\" data-starttime=\"00:23:01\">00:23:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:06\">00:23:06</a>. The optimized schema yields over 50% savings in request units for frequent reads and up to 75% savings for order retrieval operations, showing dramatic cost reductions—an estimated $980,000 per month at production scale <a class=\"medius-aisummary\" data-starttime=\"00:24:07\">00:24:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:29\">00:24:29</a>. He stresses the importance of iterative validation and simulation through the Agent Kit before deploying at full scale, ensuring database efficiency and avoiding costly schema mistakes. Marco and Sergei conclude by encouraging developers to use these AI-driven tools and the Cosmos DB emulator to continuously validate schemas and evolve NoSQL designs intelligently <a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:32\">00:25:32</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:30 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:55:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/43d42b09-edf5-442a-8016-5661a369e0f1","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/43d42b09-edf5-442a-8016-5661a369e0f1","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/43d42b09-edf5-442a-8016-5661a369e0f1","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/43d42b09-edf5-442a-8016-5661a369e0f1","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/43d42b09-edf5-442a-8016-5661a369e0f1","speakerIds":["1768245294609build26-1771861776769001JKqX","1768245294609build26-1776352718006001ATTV"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/43d42b09-edf5-442a-8016-5661a369e0f1","ogTitle":"Ship code faster with AI-powered NoSQL schema design","ogDescription":"NoSQL schema design is hard—denormalization decisions, partition key selection, and data modeling patterns require expertise. Use GitHub Copilot and the new Azure Cosmos DB Agent Toolkit to accelerate development with AI-assisted schema generation, query optimization suggestions, and refactoring recommendations. Iterate rapidly with the new Mac/Linux emulator for local testing. Demo shows schema evolution across three iterations in 30 minutes versus days of manual design.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771861776769001JKqX","RegistrantKey":"1768245294609build26-1771861776769001JKqX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776352718006001ATTV","RegistrantKey":"1768245294609build26-1776352718006001ATTV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM310","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"cc7f63c4-e450-4704-a7aa-84335846df02","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Caroline Matthews, Ryan Whitehead","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776886819724001IlSf","langLocale":"en-US","sessionId":"1776886819724001IlSf","sessionInstanceId":"1776886819724001IlSf","sessionCode":"DEMSP388","sessionCodeNormalized":"EN-US-DEMSP388","title":"Ship faster with Claude Code and Cowork in Microsoft Foundry","sortTitle":"ship faster with claude code and cowork in microsoft foundry","description":"Agentic coding has moved from research demos to production engineering. Claude can now handle multi-hour tasks across large codebases: reading existing code, planning across files, running tests, and recovering when things go wrong. Walk through what makes long-running coding agents work in practice inside Microsoft Foundry: the Claude Code harness, how it manages context, and the patterns engineers use to keep agents on track. We will build it live with Claude in Microsoft Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The presentation begins with Caroline Matthews from Anthropic greeting the audience and introducing herself as an applied AI architect there <a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>. She explains that her talk focuses on building with agentic coding agents in Microsoft Foundry and sets the stage by discussing how the landscape of software development has evolved—from traditional coding and autocomplete suggestions to fully autonomous agents capable of writing complete applications <a class=\"medius-aisummary\" data-starttime=\"00:00:21\">00:00:21</a>. Caroline provides a brief history of Anthropic’s Frontier model progression over six generations spanning roughly 18 months, highlighting performance metrics like SuiteBench for coding tasks and emphasizing how they’ve achieved near-expert accuracy levels in model-generated code <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>.</p>\n\n<p><strong>Model Advancements and Integration:</strong> She continues by elaborating on Foundry’s integration capabilities, which combine Anthropic’s safe, governed AI models with Microsoft’s enterprise stack to accelerate the path from pilot to production <a class=\"medius-aisummary\" data-starttime=\"00:01:58\">00:01:58</a>. Caroline then announces the recent model release, Opus 4.8, describing it as Anthropic’s most advanced Frontier Intelligence offering ideal for accuracy-critical, long-horizon tasks <a class=\"medius-aisummary\" data-starttime=\"00:02:12\">00:02:12</a>. She compares it with other models—Sonnet 4.6, billed as a cost-effective coding agent, and Haiku, praised for its exceptional speed and low-latency tasks such as summarization and text classification <a class=\"medius-aisummary\" data-starttime=\"00:02:56\">00:02:56</a>. Importantly, all models share Anthropic’s constitutional AI foundation, meaning consistent safety and prompt strategies across deployments.</p>\n\n<p><strong>Agentic Coding Frontiers:</strong> Caroline transitions to the main theme—exploring the “frontiers” of agentic coding, defining four conceptual challenges that represent boundaries of advancement: what agents can touch, how they validate their correctness, what they can perceive, and when/where they can operate autonomously <a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>. To illustrate these ideas, she initiates a demonstration by launching a long-running task in Cloud Code—specifically, the creation of an arcade game <a class=\"medius-aisummary\" data-starttime=\"00:05:04\">00:05:04</a>. She introduces “dynamic workflows,” a new feature in Opus 4.8 enabling modular, sub-agent collaboration. This sets the foundation for hands-off development where agents work concurrently while the user continues other activities.</p>\n\n<p><strong>First Frontier – Autonomy and Safety:</strong> Caroline discusses reducing human bottlenecks by introducing auto mode, a new classifier-based safety layer between model requests and execution permissions <a class=\"medius-aisummary\" data-starttime=\"00:06:35\">00:06:35</a>. Earlier systems required constant manual confirmations for simple tasks, but now permissions can be intelligently inferred—allowing non-risky actions to proceed automatically while flagging potentially destructive commands for review <a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>. Through a live demo, she contrasts manual approval workflows against auto mode execution, showing how developers can literally step away from their desktop and return to completed tasks safely managed by the classifier <a class=\"medius-aisummary\" data-starttime=\"00:08:44\">00:08:44</a>. This marks significant progress toward true agent autonomy in enterprise coding.</p>\n\n<p><strong>Second and Third Frontiers – Evaluating and Perceiving Code:</strong> The next section explores enabling AI agents to verify their own work. Caroline describes a multi-agent architecture composed of a planner, generator, and evaluator, emphasizing the importance of building strong evaluation agents first <a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>. She demonstrates this with an “Oracle” reviewer setup, where one agent writes a simple function and another critiques it through test cases, effectively serving as a tough code reviewer <a class=\"medius-aisummary\" data-starttime=\"00:11:11\">00:11:11</a>. This evaluative separation is crucial for scaling quality control. Moving to perception, Caroline showcases how Claude Opus can now use browser-based vision and high-resolution image processing to perform tasks such as penetration testing <a class=\"medius-aisummary\" data-starttime=\"00:13:06\">00:13:06</a>. In this demo, the AI identifies hidden application security risks from a screenshot far more effectively than humans, demonstrating its advanced reasoning capabilities over images <a class=\"medius-aisummary\" data-starttime=\"00:15:20\">00:15:20</a>.</p>\n\n<p><strong>Fourth Frontier and Closing Insights:</strong> The final frontier concerns when and where agents can run. Caroline describes continuous, multi-surface execution—agents capable of running across devices, looping through schedules, and completing tasks in parallel even while developers attend meetings <a class=\"medius-aisummary\" data-starttime=\"00:17:01\">00:17:01</a>. She revisits the earlier arcade game demo, revealing it successfully completed using dynamic workflows that orchestrated several sub-agents who designed, judged, and tested different game versions <a class=\"medius-aisummary\" data-starttime=\"00:18:08\">00:18:08</a>. The result, “Neon Ascent,” exemplifies how autonomous yet structured agents collaborate without human micromanagement. Caroline concludes by emphasizing that the goal isn’t to replace humans but to elevate their focus—leaving repetitive code tasks to agentic systems while humans guide critical, high-value decisions <a class=\"medius-aisummary\" data-starttime=\"00:20:40\">00:20:40</a>. The talk closes with appreciation and encouragement to continue advancing agentic coding responsibly within safe AI boundaries <a class=\"medius-aisummary\" data-starttime=\"00:21:07\">00:21:07</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:10 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:35:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/46e9fb81-4bb9-4d1c-a5e4-ca32bd1bb8ff","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/46e9fb81-4bb9-4d1c-a5e4-ca32bd1bb8ff","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/46e9fb81-4bb9-4d1c-a5e4-ca32bd1bb8ff","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/46e9fb81-4bb9-4d1c-a5e4-ca32bd1bb8ff","speakerIds":["1768245294609build26-1778020065215001t4NN","1768245294609build26-1777992712377001cayS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.386Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Anthropic","PBC"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/46e9fb81-4bb9-4d1c-a5e4-ca32bd1bb8ff","ogTitle":"Ship faster with Claude Code and Cowork in Microsoft Foundry","ogDescription":"Agentic coding has moved from research demos to production engineering. Claude can now handle multi-hour tasks across large codebases: reading existing code, planning across files, running tests, and recovering when things go wrong. Walk through what makes long-running coding agents work in practice inside Microsoft Foundry: the Claude Code harness, how it manages context, and the patterns engineers use to keep agents on track. We will build it live with Claude in Microsoft Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777992712377001cayS","RegistrantKey":"1768245294609build26-1777992712377001cayS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778020065215001t4NN","RegistrantKey":"1768245294609build26-1778020065215001t4NN"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Building%20with%20Agenting%20Coding%20Agents_MSFT%20Build%20Edit_1779543693087001lBVV_1780514531364001N2cX.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"0d0ddf54-6da9-459d-8138-4c2caa663019","startDateTime":"2026-06-03T18:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maxim Lukiyanov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776289985706001LtgP","langLocale":"en-US","sessionId":"1776289985706001LtgP","sessionInstanceId":"1776289985706001LtgP","sessionCode":"DEM364","sessionCodeNormalized":"EN-US-DEM364","title":"Simplify app dev with cloud-native PostgreSQL in Azure HorizonDB","sortTitle":"simplify app dev with cloud-native postgresql in azure horizondb","description":"Enterprise developers creating AI‑driven applications face growing complexity as vector search, models, and retrieval pipelines sprawl across services. In this session, see how Azure HorizonDB streamlines the stack by embedding AI and search directly in the database. Learn to run hybrid vector queries, apply BM25 relevance, call managed AI models from SQL, and prototype agentic workflows—shipping faster with less architectural overhead.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session begins with a warm introduction at <a class=\"medius-aisummary\" data-starttime=\"00:00:22\">00:00:22</a>, welcoming viewers to a demo titled “Simplify app development with cloud native PostgreSQL in Azure Horizon DB.” The presenter provides context by referencing Satya’s recent announcement about Azure Horizon DB entering public preview <a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>. Horizon DB is introduced as a new cloud-optimized PostgreSQL database that connects the open-source engine with scalable storage to achieve high resiliency, larger scale, and improved performance. The speaker highlights that Horizon DB delivers three times higher throughput and vector search speed while supporting up to fifteen read-only replicas for enhanced scalability <a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>. Importantly, the demo will focus on how its built-in AI models and pipelines simplify app development by constructing a live example on stage.</p>\n\n<p><strong>Demo Application Setup:</strong> The demonstration transitions at <a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a> to an example project—a “Zava room designer” app intended to automatically suggest furniture that matches a user’s room style based on a photograph. The presenter shows how the app can process an image, identify design features, and recommend compatible furniture <a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>. Although the application appears complex, it is deliberately designed to be simple. The following portion of the demo walks through how each component is built incrementally. The first stage, beginning at <a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>, demonstrates provisioning a Horizon DB instance directly through the Azure portal and enabling AI features via a simple checkbox that includes model management, vector search, and PG extensions. The setup section concludes with a ready-to-use pre-provisioned instance that already has embedded AI models for immediate experimentation.</p>\n\n<p><strong>Data Preparation and AI Pipelines:</strong> With the database provisioned, the presenter at <a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a> begins preparing product catalog data for AI consumption by generating vector embeddings. The demo shows a SQL-based workflow that inspects the model registry, revealing prepackaged models including an embedding model, chat completion model, and semantic ranker <a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>. Next, an AI pipeline is defined entirely in SQL to chunk, embed, and store data directly within Horizon DB. Once executed, the system asynchronously creates vectorized outputs and updates them automatically whenever new catalog entries are added. Verified through the UI at <a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>, the encryption of product data into 116 vectors demonstrates Horizon DB’s ability to maintain an up-to-date semantic representation of information, eliminating the need for separate external feature processing services. The end result is a fully vectorized catalog that can be queried using standard SQL syntax to locate relevant items such as \"comfortable chairs\" <a class=\"medius-aisummary\" data-starttime=\"00:12:23\">00:12:23</a>.</p>\n\n<p><strong>Search Enhancement and Semantic Ranking:</strong> Building upon the prepared data, the next segment introduces advanced retrieval methods beginning at <a class=\"medius-aisummary\" data-starttime=\"00:14:16\">00:14:16</a>. Horizon DB supports hybrid search by combining full-text search (through the new PGFTS extension) with fast vector indexing for extremely efficient query performance. A single hybrid function allows developers to retrieve results that are both semantically and textually relevant. When semantic re-ranking is enabled at <a class=\"medius-aisummary\" data-starttime=\"00:17:31\">00:17:31</a>, it improves query accuracy using the embedded ranker model to reorder search results by real meaning rather than keyword frequency. The presenter visualizes the underlying query plan at <a class=\"medius-aisummary\" data-starttime=\"00:19:05\">00:19:05</a>, revealing that hybrid search runs textual and vector queries in parallel, merges ranking lists using a reciprocal rank fusion method, then applies semantic reordering for optimal results. This powerful orchestration demonstrates how sophisticated search intelligence can be achieved directly inside PostgreSQL with minimal configuration.</p>\n\n<p><strong>Knowledge Graphs and Contextual Intelligence:</strong> The final technical demonstration begins around <a class=\"medius-aisummary\" data-starttime=\"00:21:02\">00:21:02</a> by showcasing the creation of a style-based knowledge graph using the Apache AGE extension integrated within Horizon DB. This feature represents relationships between furniture items and styles such as “mid-century modern,” connecting each to corresponding categories like chairs or tables <a class=\"medius-aisummary\" data-starttime=\"00:23:01\">00:23:01</a>. The visualized graph also models similarities between different styles—such as linking mid-century modern and Bahamian—allowing broader recommendations even when inventories vary. The presenter emphasizes that knowledge graphs store domain-specific logic in graph form while still residing in PostgreSQL, thereby augmenting AI agents with deep contextual understanding of data relationships.</p>\n\n<p><strong>Conclusion and Recap:</strong> Concluding at <a class=\"medius-aisummary\" data-starttime=\"00:24:40\">00:24:40</a>, the session reviews how Horizon DB enables developers to integrate AI pipelines, perform real-time hybrid search, and enrich results through semantic ranking and knowledge graphs—all within a single managed PostgreSQL environment. The speaker reiterates that all functionality demonstrated was powered by built-in Azure Horizon DB AI features without external services or complex orchestration. Viewers are encouraged to explore the public code repository containing sample datasets, queries, and scripts <a class=\"medius-aisummary\" data-starttime=\"00:25:45\">00:25:45</a>. The demo closes by highlighting how Horizon DB drastically simplifies intelligent app development while maintaining enterprise-grade scalability and performance.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"18:20 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/17b83627-30f4-4447-9093-6446eef5d675","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/17b83627-30f4-4447-9093-6446eef5d675","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/17b83627-30f4-4447-9093-6446eef5d675","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/17b83627-30f4-4447-9093-6446eef5d675","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/17b83627-30f4-4447-9093-6446eef5d675","speakerIds":["1768245294609build26-1765837200716001tuxr"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.903Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/17b83627-30f4-4447-9093-6446eef5d675","ogTitle":"Simplify app dev with cloud-native PostgreSQL in Azure HorizonDB","ogDescription":"Enterprise developers creating AI‑driven applications face growing complexity as vector search, models, and retrieval pipelines sprawl across services. In this session, see how Azure HorizonDB streamlines the stack by embedding AI and search directly in the database. Learn to run hybrid vector queries, apply BM25 relevance, call managed AI models from SQL, and prototype agentic workflows—shipping faster with less architectural overhead.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765837200716001tuxr","RegistrantKey":"1768245294609build26-1765837200716001tuxr"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM364","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"b46fe6b8-64d6-47cd-bfa3-4c65066faa59","startDateTime":"2026-06-03T19:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"William Liang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018714376001r0Qn","langLocale":"en-US","sessionId":"1774018714376001r0Qn","sessionInstanceId":"1774018714376001r0Qn","sessionCode":"DEM322","sessionCodeNormalized":"EN-US-DEM322","title":"Smaller, faster, smarter: Distilling models with fine‑tuning","sortTitle":"smaller, faster, smarter: distilling models with fine‑tuning","description":"Large models are powerful, but expensive to run in production. In this demo, we’ll show how teams use Foundry for distillation and supervised fine-tuning to train small language models for task‑specific accuracy, dramatically reducing latency and cost. We’ll cover when distillation makes sense, how it complements fine tuning and reinforcement learning, and what real production teams have learned when deploying smaller models at scale. Expect fast examples and lots of Q&A.","aiDescription":"<p><strong>Introduction and Motivation:</strong> The session opens with greetings and a brief overview of the purpose of the talk <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>. William from the Foundry Fine Tuning team explains that six months ago, the focus was on building the fastest agent, but now the objective is shifting toward affordability and scalability in enterprise contexts <a class=\"medius-aisummary\" data-starttime=\"00:00:14\">00:00:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:30\">00:00:30</a>. As AI transitions from research labs into enterprise workflows, it must become cheaper and more accessible—a utility rather than a luxury. The audience is told the session will demonstrate how production agent traces can be transformed into smaller, cost-effective models <a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>.</p>\n\n<p><strong>Understanding Model Distillation and Agent Traces:</strong> William introduces the core concept of model distillation, describing it as an apprenticeship between teacher and student models <a class=\"medius-aisummary\" data-starttime=\"00:01:37\">00:01:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:12\">00:02:12</a>. A larger model performs tasks in production, generating examples that guide the smaller model's fine-tuning process. He further explains why learning from agent traces—real user interactions—is valuable for training models that replicate realistic problem-solving behavior <a class=\"medius-aisummary\" data-starttime=\"00:02:24\">00:02:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:05\">00:03:05</a>. Real traces allow models to learn not just dialogue, but which tools to use, in what order, and with what arguments, teaching them correct decision-making trajectories. Foundry provides built-in tools to collect these traces automatically, allowing agents to grow smarter through continued interaction <a class=\"medius-aisummary\" data-starttime=\"00:03:37\">00:03:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>.</p>\n\n<p><strong>Distillation Benefits and Setup for Demo:</strong> The presenter outlines four wins from distilling models: lower cost, faster responses, quality comparable to teacher models, and improved behavioral consistency through supervised fine-tuning <a class=\"medius-aisummary\" data-starttime=\"00:04:23\">00:04:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:12\">00:05:12</a>. To demonstrate this practically, William transitions to a live demo, explaining that difficult tasks will be given to both teacher (GPT‑55) and student (GPT‑41 Nano) models <a class=\"medius-aisummary\" data-starttime=\"00:05:26\">00:05:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:11\">00:06:11</a>. The teacher model will generate rich traces simulating real-world interactions, which will then be cleaned and used to fine-tune the student model. The setup emphasizes how costly teacher models can be, motivating the need for efficient distillation strategies to produce smaller models that maintain accuracy but operate faster and cheaper <a class=\"medius-aisummary\" data-starttime=\"00:06:44\">00:06:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:02\">00:07:02</a>.</p>\n\n<p><strong>Evaluation Metrics and Fine-Tuning Process:</strong> After simulating tasks, William discusses evaluating model performance through multi-dimensional scoring, including decision correctness, trajectory accuracy, financial accuracy, and communication quality <a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:50\">00:09:50</a>. He introduces the \"pass K\" metric—how often an agent succeeds above a quality threshold after repeated trials <a class=\"medius-aisummary\" data-starttime=\"00:10:34\">00:10:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:14\">00:11:14</a>. The results show that the fine-tuned student narrows the capability gap dramatically compared to its teacher <a class=\"medius-aisummary\" data-starttime=\"00:11:24\">00:11:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:04\">00:12:04</a>. Visual comparisons confirm significant performance lift, proving distillation’s effectiveness. The workflow is then demonstrated on Foundry, showing how traces collected automatically from real conversations can be converted into supervised fine-tuning datasets, ensuring scalability and enterprise integration <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:09\">00:16:09</a>.</p>\n\n<p><strong>Real-World Results and Case Examples:</strong> William shares several concrete case studies to highlight performance improvements. One example involves a refund scenario where the base model misclassified a customer complaint, while the fine-tuned version correctly applied business policy and prevented unnecessary refunds <a class=\"medius-aisummary\" data-starttime=\"00:17:08\">00:17:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>. Another example shows improved procedural compliance—after tuning, the agent checks the fulfillment policy before issuing refunds, reducing risk <a class=\"medius-aisummary\" data-starttime=\"00:19:24\">00:19:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:17\">00:21:17</a>. Further examples involve consistent handling of defective-item claims and correctly rejecting out-of-scope requests such as new orders <a class=\"medius-aisummary\" data-starttime=\"00:21:22\">00:21:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:00\">00:24:00</a>. Each demonstrates improved reasoning, compliance with business logic, and more confidence in automated decision-making due to the distilled agent’s learning from teacher traces.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> The session closes with a reflection on scaling intelligence affordably within organizations <a class=\"medius-aisummary\" data-starttime=\"00:24:11\">00:24:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:05\">00:25:05</a>. The key message emphasizes treating intelligence as an embedded enterprise capability, not a luxury. By leveraging existing agent traces and production data, companies can distill expensive frontier models into tailored, efficient agents suited for specific business needs. William thanks attendees and expresses hope that they will apply these techniques to create smaller, faster, and smarter models that democratize AI within their environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"19:20 - 19:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:45:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","speakerIds":["1768245294609build26-1774284998627001p6H4"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.231Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/47728171-6ede-46aa-9dd5-63ec9ac0ffbb","ogTitle":"Smaller, faster, smarter: Distilling models with fine‑tuning","ogDescription":"Large models are powerful, but expensive to run in production. In this demo, we’ll show how teams use Foundry for distillation and supervised fine-tuning to train small language models for task‑specific accuracy, dramatically reducing latency and cost. We’ll cover when distillation makes sense, how it complements fine tuning and reinforcement learning, and what real production teams have learned when deploying smaller models at scale. Expect fast examples and lots of Q&A.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998627001p6H4","RegistrantKey":"1768245294609build26-1774284998627001p6H4"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"700982ea-2e9d-402d-8396-6ffc3d31b733","startDateTime":"2026-06-02T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Christine Flora, Fernanda Herger, Stephen Simon, Jeremy Sinclair, Betsy Weber","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774042138922001oQn9","langLocale":"en-US","sessionId":"1774042138922001oQn9","sessionInstanceId":"1774042138922001oQn9","sessionCode":"DEM375","sessionCodeNormalized":"EN-US-DEM375","title":"So you want to become an MVP?","sortTitle":"so you want to become an mvp?","description":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introductions and Overview of the MVP Program:</strong> At the start of the video (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>), Betsy Weber, Senior Engagement Lead at Microsoft, welcomes the audience and introduces Fernanda Sariva, Principal Customer Experience Program Manager. They begin by explaining the Microsoft MVP (Most Valuable Professional) Program (<a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>). The MVP award recognizes outstanding community leaders who freely share technical expertise, provide real-world insights on Microsoft products, and deliver feedback to help shape innovations. Betsy emphasizes that MVPs are passionate contributors who actively engage with and help build the Microsoft community worldwide.</p>\n\n<p><strong>Global Reach and Community Impact:</strong> Fernanda continues by describing the global landscape of the MVP program (<a class=\"medius-aisummary\" data-starttime=\"00:01:32\">00:01:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:32\">00:02:32</a>). With over 4,000 influencers across 96 countries speaking nearly 50 languages, the MVP network is a diverse and vibrant community. She encourages viewers who are active in local tech communities to consider joining, noting the continual expansion of the program’s global influence. This section highlights the collaborative spirit that defines MVPs—building communities, fostering knowledge exchange, and promoting growth through collective engagement.</p>\n\n<p><strong>Becoming an MVP and Areas of Contribution:</strong> The conversation then focuses on how individuals can become MVPs (<a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:28\">00:05:28</a>). Fernanda explains that potential MVPs should start by engaging in activities they are passionate about—such as writing blogs, creating videos, contributing to open-source projects, or supporting peer groups. Technical expertise, leadership, and meaningful community impact over time are key criteria. She emphasizes that contributions are not limited to Microsoft technologies; people who love sharing knowledge, regardless of their technical focus, are welcome. Betsy clarifies that blogging, podcasting, or open-source work are optional ways to participate—the heart of the program lies in genuine community involvement and sharing expertise freely.</p>\n\n<p><strong>Benefits and Recognition of MVPs:</strong> Betsy outlines the numerous benefits and experiences MVPs enjoy (<a class=\"medius-aisummary\" data-starttime=\"00:06:56\">00:06:56</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:32\">00:08:32</a>). These include professional recognition, access to non-disclosure (NDA) product insights, roadmaps from Microsoft product groups, and feedback opportunities for shaping future development. MVPs also attend exclusive events such as the annual MVP Global Summit at Redmond, gaining chances for collaboration and networking with engineers and other community leaders. The experience offers personal and professional growth, credibility within the tech community, and connections worldwide.</p>\n\n<p><strong>Real MVP Stories and Guidance:</strong> The session transitions to interviews with real MVPs like Jeremy Sinclair and Christine (<a class=\"medius-aisummary\" data-starttime=\"00:09:30\">00:09:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:56\">00:24:56</a>). They share how they first got involved, the personal and career transformations that followed, and strategies for balancing work and community contributions. Their advice centers on authenticity—doing what you love, staying consistent, and helping others without focusing solely on earning the award. Betsy reinforces this point by noting that the most re-awarded MVPs are those who contribute continuously out of passion rather than obligation.</p>\n\n<p><strong>Wrap-Up and How to Get Nominated:</strong> In conclusion (<a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:25:47\">00:25:47</a>), Betsy and Fernanda summarize three simple steps to become an MVP: be an expert, do what you love, and let Microsoft know about your contributions. Nominations can be made by either a current MVP or a Microsoft employee. They encourage the audience to visit the MVP Learn website (aka.ms/mvp/learn) to explore resources and opportunities, inviting continued engagement across future sessions. The video ends on an encouraging note—celebrating community, collaboration, and the ever-growing impact of the Microsoft MVP ecosystem.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:00 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/627da16b-571e-42f2-b209-04848e325852","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/627da16b-571e-42f2-b209-04848e325852","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/627da16b-571e-42f2-b209-04848e325852","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/627da16b-571e-42f2-b209-04848e325852","speakerIds":["1768245294609build26-1772578847951001DH4K","1768245294609build26-1776209926960001eLK3","1768245294609build26-1774323608465001D4tS","1768245294609build26-1772566305259001tYa6","1768245294609build26-1758299331835001JwKs"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.633Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/627da16b-571e-42f2-b209-04848e325852","ogTitle":"So you want to become an MVP?","ogDescription":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758299331835001JwKs","RegistrantKey":"1768245294609build26-1758299331835001JwKs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776209926960001eLK3","RegistrantKey":"1768245294609build26-1776209926960001eLK3"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772566305259001tYa6","RegistrantKey":"1768245294609build26-1772566305259001tYa6"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774323608465001D4tS","RegistrantKey":"1768245294609build26-1774323608465001D4tS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772578847951001DH4K","RegistrantKey":"1768245294609build26-1772578847951001DH4K"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"33743da5-e9f8-4e9c-a02a-253ec90863fe","startDateTime":"2026-06-03T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Shannon Eldridge-Kuehn, Fernanda Herger, Stephen Simon, Jeremy Sinclair, Betsy Weber","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778799822091001NW9j","langLocale":"en-US","sessionId":"1778799822091001NW9j","sessionInstanceId":"1778799822091001NW9j","sessionCode":"DEM375-R1","sessionCodeNormalized":"EN-US-DEM375-R1","title":"So you want to become an MVP?","sortTitle":"so you want to become an mvp?","description":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Presenter Backgrounds:</strong> The session opens with an introduction and greetings from the hosts Betsy Weber and Fernanda Sariva <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:15\">00:00:15</a>. Betsy, celebrating her ten years at Microsoft, identifies herself as a Senior Engagement Lead on the Microsoft MVP Program, while Fernanda shares that she has spent over fifteen years at the company as a Principal Customer Experience Manager, previously also an MVP. Together, they outline that the event’s focus is to share insights about the MVP community—what the program is, how members participate, and how they engage globally. The introduction establishes their credibility and connection to the MVP ecosystem and previews that they will bring in MVPs to share firsthand experiences.</p>\n\n<p><strong>Understanding the MVP Program:</strong> Betsy explains the essence of the Microsoft MVP recognition <a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:42\">00:01:42</a>. The MVP (Most Valuable Professional) title is awarded to impactful community members who contribute by teaching others, writing technical books, organizing events, or developing materials about Microsoft technologies. Fernanda adds historical context—the program has existed for over thirty years, now comprising more than 4,000 active MVPs across 96 countries representing 49 languages <a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:24\">00:02:24</a>. She outlines that MVPs focus on 112 technology areas grouped into award categories such as M365 or developer technologies. Their contributions demonstrate personal expertise, community engagement, and authentic knowledge sharing rather than promotional work. Fernanda describes four key pillars that define MVP selection: technical expertise, product feedback, impact on the Microsoft ecosystem, and community health <a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:22\">00:05:22</a>.</p>\n\n<p><strong>Nomination Process and MVP Benefits:</strong> Betsy clarifies the misconceptions around qualifying as an MVP—there is no rigid formula requiring blogs, books, or videos; contributions should reflect authentic passion <a class=\"medius-aisummary\" data-starttime=\"00:05:49\">00:05:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:39\">00:06:39</a>. MVPs are community leaders who volunteer their time outside their paid roles and are nominated by a Microsoft employee or an existing MVP based on the previous year’s achievements <a class=\"medius-aisummary\" data-starttime=\"00:07:02\">00:07:02</a>. She elaborates that honorees receive unique opportunities—recognition, access to insider NDA information, product updates, early previews of new Microsoft technologies, and invitations to exclusive events like Microsoft Build <a class=\"medius-aisummary\" data-starttime=\"00:08:06\">00:08:06</a>. The MVP status builds connections with global peers and fosters collaboration with product groups. Fernanda and Betsy highlight the cross-cultural and multilingual reach of the program, reinforcing that it thrives on diversity and curiosity rather than prestige alone.</p>\n\n<p><strong>Personal MVP Journeys and Professional Impact:</strong> Two guest MVPs—Shannon Keen and Steven Simon—share their experiences <a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>. Shannon, based in Chicago, recounts her transition from a Microsoft employee to an MVP after contributing steadily to the community through video content and advocacy. Steven, from Delhi, traces his journey from Microsoft Student Partner to MVP after years of community involvement. Both attribute the award to steady dedication rather than a single event. Shannon explains how being an MVP sparked valuable client conversations and opened professional doors <a class=\"medius-aisummary\" data-starttime=\"00:15:01\">00:15:01</a>, while Steven highlights the credibility boost and professional growth—his MVP page enhanced his personal branding and led to new speaking and job opportunities <a class=\"medius-aisummary\" data-starttime=\"00:13:40\">00:13:40</a>. They agree that mentorship, networking, and sharing knowledge are core benefits that amplify both career and community impact.</p>\n\n<p><strong>Advice for Aspiring Community Contributors:</strong> The MVPs offer guidance for those starting in community involvement <a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:11\">00:17:11</a>. Shannon encourages identifying a passion area—whether writing, coding, or presenting—and connecting with others through collaborative ideas. She emphasizes using mentorship to nurture skills and confidence. Steven advises reaching out to existing MVPs or organizers with a willingness to help, as community involvement grows from engagement and initiative. Fernanda and Betsy reinforce that contribution shouldn’t feel burdensome; rather, it should reflect natural curiosity and enjoyment of learning. The discussion evolves into how MVPs balance community work with professional life <a class=\"medius-aisummary\" data-starttime=\"00:19:40\">00:19:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:20\">00:21:20</a>. Shannon shares time-management tips by using calendar focus blocks, while Steven humorously adds flexibility and passion as his balance strategy. Together they underscore that sustainable community participation grows from genuine interest.</p>\n\n<p><strong>Conclusion and Path to Becoming an MVP:</strong> The closing remarks summarize the essence of becoming an MVP—being an expert who naturally contributes to the community out of enthusiasm, not obligation <a class=\"medius-aisummary\" data-starttime=\"00:22:20\">00:22:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:43\">00:23:43</a>. Betsy and Fernanda remind viewers that MVPs are nominated based on their previous year’s work and judged holistically on impact and authenticity. The simple “secret sauce” is continuous learning and generous sharing—posting blogs, videos, how-tos, or simply lessons learned publicly. They close with encouragement that aspiring contributors should find their niche, connect with others, and express their technical curiosity at any level—from beginner to advanced. Finally, they invite the audience to explore more resources and official details about the MVP program at the Microsoft MVP learning portal (aka.ms/mvp/learn) <a class=\"medius-aisummary\" data-starttime=\"00:24:35\">00:24:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:53\">00:24:53</a>, thanking attendees for their time and inspiring them to join and grow within the vibrant global community.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:30 - 22:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a4918e84-1b84-4309-a75e-d27409bc3d70","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a4918e84-1b84-4309-a75e-d27409bc3d70","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a4918e84-1b84-4309-a75e-d27409bc3d70","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a4918e84-1b84-4309-a75e-d27409bc3d70","speakerIds":["1768245294609build26-1770655732723001OM3q","1768245294609build26-1776209926960001eLK3","1768245294609build26-1774323608465001D4tS","1768245294609build26-1772566305259001tYa6","1768245294609build26-1758299331835001JwKs"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.233Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a4918e84-1b84-4309-a75e-d27409bc3d70","ogTitle":"So you want to become an MVP?","ogDescription":"Passionate about sharing knowledge and making an impact in tech? Learn what it takes to become a Microsoft MVP from those who’ve done it and hear directly from MVP Program staff. Discover how to expand your influence, showcase your expertise, and practical tips for getting nominated. Walk away with a roadmap for growing your contributions and positioning yourself for MVP success.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758299331835001JwKs","RegistrantKey":"1768245294609build26-1758299331835001JwKs"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776209926960001eLK3","RegistrantKey":"1768245294609build26-1776209926960001eLK3"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770655732723001OM3q","RegistrantKey":"1768245294609build26-1770655732723001OM3q"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774323608465001D4tS","RegistrantKey":"1768245294609build26-1774323608465001D4tS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772566305259001tYa6","RegistrantKey":"1768245294609build26-1772566305259001tYa6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"c5158cc4-6443-4658-b3e1-e1190e239073","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nik Karpinsky","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018709769001rxEg","langLocale":"en-US","sessionId":"1774018709769001rxEg","sessionInstanceId":"1774018709769001rxEg","sessionCode":"DEM300","sessionCodeNormalized":"EN-US-DEM300","title":"Supercharged profiling: Finding performance bugs with agents","sortTitle":"supercharged profiling: finding performance bugs with agents","description":"Modern performance problems are harder than ever to diagnose: distributed systems, async code, and massive data make traditional profiling insufficient. In this demo-driven session, watch how Visual Studio combines advanced diagnostics with AI-powered agents to surface bottlenecks, explain root causes, and guide fixes faster than ever. From CPU and memory issues to “it only fails in production” bugs, you’ll see how profiling evolves in an agentic world.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The session begins with Nick Karpinski, a Microsoft software engineer with over twelve years of experience on Visual Studio, opening the first demo of the morning <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>. He introduces himself and explains his focus on the Visual Studio Profiler, expressing excitement to share how developers can optimize workflows using the profiler in Visual Studio. Before diving into technical content, he engages the audience to gauge their experience levels with .NET, Visual Studio Profiler, and Benchmark.NET <a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>. Noting that many attendees are unfamiliar with these tools, he sets the stage for a demo-heavy presentation focused on demonstrating the optimization process in practice.</p>\n\n<p><strong>Performance Optimization Workflow:</strong> Nick outlines a high-level flowchart for real-world code optimization <a class=\"medius-aisummary\" data-starttime=\"00:01:27\">00:01:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:02\">00:04:02</a>, describing steps like understanding performance issues, using profiler data to identify bottlenecks, and applying changes iteratively. He stresses the importance of measuring performance with concrete data rather than intuition, recommending developers start by profiling their applications to pinpoint time-consuming areas. Once bottlenecks are known, he introduces Benchmark.NET as a reusable performance testing framework, comparing it to unit tests that validate functionality but focused instead on measuring code efficiency. He encourages developers to adopt the “measure twice, optimize once” principle: collect baseline data, apply code changes, and remeasure to understand impact. Optimizations are treated as iterative experiments, looping through improvements until performance gains plateau or new issues arise.</p>\n\n<p><strong>Benchmark and Demo Setup:</strong> Transitioning to the live demo, Nick opens the CSV Helper repository—a popular NuGet package—illustrating how to approach optimizing community code <a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:04\">00:07:04</a>. He explains that CSV Helper already includes a Benchmark.NET project for testing read performance, which he previously contributed. He reviews how benchmarks isolate targeted components of code for focused measurement, using memory streams to avoid unrelated IO overhead. After establishing a baseline for reading CSV files, he decides to create a complementary benchmark for writing CSV files <a class=\"medius-aisummary\" data-starttime=\"00:07:07\">00:07:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:55\">00:09:55</a>. Using Copilot Chat connected to the profiler agent, Nick instructs the system to generate a new benchmark that mirrors the existing read test. Copilot automatically reads the code context, installs necessary integration packages for Benchmark.NET and Visual Studio Profiling, ensuring diagnostics are captured during runtime.</p>\n\n<p><strong>Profiling, Measurement, and Optimization:</strong> Once the benchmark compiles successfully, Nick runs the “write records” benchmark through Copilot’s profiler integration <a class=\"medius-aisummary\" data-starttime=\"00:10:16\">00:10:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a>. The profiler collects performance traces and highlights the main CPU hotspots such as “expression compile” and “should quote” logic. To demonstrate iterative optimization, he focuses on improving the “should quote” method rather than the top CPU consumer, using Copilot to analyze source code lines linked to performance overhead via the profiler’s “go to source” functionality. Copilot suggests combining multiple character checks into a single iteration to reduce redundant scans <a class=\"medius-aisummary\" data-starttime=\"00:14:06\">00:14:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:03\">00:15:03</a>. After applying this optimization, Nick remeasures benchmark results, noting that while overall runtime remained near 1.4 milliseconds, specific sections showed reduced CPU consumption—from 13% down to 7%. This highlights that measurable data, not assumptions, defines true optimization outcomes.</p>\n\n<p><strong>Iterative Improvements and Learning:</strong> Nick emphasizes using Copilot and the profiler not only for automation but for guided learning <a class=\"medius-aisummary\" data-starttime=\"00:16:02\">00:16:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:32\">00:18:32</a>. He demonstrates how engineers can ask Copilot contextual questions to understand why certain optimizations didn’t yield expected results—for instance, learning about .NET JIT quirks affecting method inlining limits. Despite minor gains in the live demo, prior tests achieved meaningful 25% improvements, showing how small incremental changes compound over time. Developers can leverage this workflow to conduct deeper investigations, validate optimization impact, and use AI assistance to interpret complex performance patterns within codebases.</p>\n\n<p><strong>Conclusion and Q&A Session:</strong> Wrapping up, Nick encourages participants to reproduce similar performance improvements using CSV Helper benchmarks available in the GitHub repository <a class=\"medius-aisummary\" data-starttime=\"00:18:33\">00:18:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:39\">00:19:39</a>. He points to related resources, including documentation for the profiler agent and a follow-up breakout session at 4:00 PM featuring debugging and unit testing integration within Visual Studio. During audience questions <a class=\"medius-aisummary\" data-starttime=\"00:20:01\">00:20:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:08\">00:23:08</a>, he explains how Copilot can inject functional tests post-optimization to ensure nothing breaks and how the profiler can interpret traces automatically using top insights that flag known inefficient patterns. These insights may include poor data structure usage or redundant operations, allowing automated guidance toward better performance solutions. Nick closes with appreciation for attendees, reinforcing the combined power of AI-assisted profiling and developer-driven analysis in practical performance engineering.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"16:30 - 16:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:55:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/fdb6bddc-1932-46a2-9662-01dc3f87fd64","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/fdb6bddc-1932-46a2-9662-01dc3f87fd64","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/fdb6bddc-1932-46a2-9662-01dc3f87fd64","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/fdb6bddc-1932-46a2-9662-01dc3f87fd64","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/fdb6bddc-1932-46a2-9662-01dc3f87fd64","speakerIds":["1768245294609build26-1774645705844001UesE"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.466Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/fdb6bddc-1932-46a2-9662-01dc3f87fd64","ogTitle":"Supercharged profiling: Finding performance bugs with agents","ogDescription":"Modern performance problems are harder than ever to diagnose: distributed systems, async code, and massive data make traditional profiling insufficient. In this demo-driven session, watch how Visual Studio combines advanced diagnostics with AI-powered agents to surface bottlenecks, explain root causes, and guide fixes faster than ever. From CPU and memory issues to “it only fails in production” bugs, you’ll see how profiling evolves in an agentic world.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774645705844001UesE","RegistrantKey":"1768245294609build26-1774645705844001UesE"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM300","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"925ec259-6d28-4299-a388-40062110cdce","startDateTime":"2026-06-03T21:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Deepti Dheer, Mike Richter","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776886897903001IMT6","langLocale":"en-US","sessionId":"1776886897903001IMT6","sessionInstanceId":"1776886897903001IMT6","sessionCode":"DEMSP395","sessionCodeNormalized":"EN-US-DEMSP395","title":"Teach AI agents to manage context with Elastic Agent Builder","sortTitle":"teach ai agents to manage context with elastic agent builder","description":"Solve AI context limits by enabling agents to manage their own memory. Learn how to prevent bloated prompts and context drift during long tasks, reduce input token usage while ensuring enterprise data governance and scalability. Walk away with next steps for deploying the dynamically loaded skills, conversation context store, selective compaction, and secure external data connectors from Elasticsearch 9.4's Agent Builder  in the Microsoft ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Microsoft–Elastic Collaboration:</strong> The session opens with Mike Richter, Principal Partner Solution Architect at Microsoft, introducing himself and the topic of discussion—Elastic running on Azure (<a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>). He highlights the strong partnership between Microsoft and Elastic, emphasizing how his role involves helping partners build and sell solutions on Azure. Mike notes that Elastic exemplifies an ideal partner for Microsoft’s cloud ecosystem, offering simplicity in procurement and enterprise-level integration. Deepti, the Product Manager from Elastic, briefly introduces herself as well, setting up the transition to the main conversation about Elastic’s cloud offering within Azure.</p>\n\n<p><strong>Elastic on Azure: Deployment and Integration Overview:</strong> Mike begins by explaining how straightforward it is for customers to purchase and deploy Elastic from the Azure Marketplace (<a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:36\">00:01:36</a>). He outlines that Elastic is currently available in around fifteen Azure regions and continually expanding. Customers with Azure commitment dollars (MAC agreements) can use those funds to procure Elastic services, simplifying enterprise billing through a single invoice. He also mentions the availability of Elastic’s hosted and serverless offerings and the benefits of zero egress costs when workloads communicate within the same Azure region (<a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:13\">00:02:13</a>). Integration features include Azure Active Directory-based authentication, private link support for secure communications, and enterprise-ready compliance configurations—all ensuring seamless data flows within customers’ virtual networks.</p>\n\n<p><strong>Elastic Cloud, Vector Support, and AI Capabilities:</strong> Mike continues by detailing Elastic’s advanced technical capabilities such as first-class vector support, making it well-suited for Retrieval-Augmented Generation (RAG) and agentic applications inside Azure (<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:29\">00:03:29</a>). He emphasizes that Microsoft Foundry’s large language models (LLMs)—including ones offered by Anthropic and OpenAI—can be seamlessly integrated to power Elastic-based agents. He transitions to mentioning Elastic’s Agent Builder, which Deepti will cover in detail, and invites attendees to visit the Microsoft booth after the session for hands-on demos and discussions about applications and agents built with Elastic’s ecosystem (<a class=\"medius-aisummary\" data-starttime=\"00:03:43\">00:03:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:08\">00:04:08</a>).</p>\n\n<p><strong>Context Gap and Architecture of Elastic Agent Builder:</strong> When Deepti takes over (<a class=\"medius-aisummary\" data-starttime=\"00:04:12\">00:04:12</a>), she starts by outlining the core problem Elastic seeks to solve—the fragmentation caused by data silos across disparate systems. She explains that AI systems fail to perform optimally when data resides in isolated contexts, so Elastic’s goal is to bridge this “context gap.” The platform achieves this by wrapping diverse data sources—both within and external to Elastic—with contextual retrieval and token optimization processes (<a class=\"medius-aisummary\" data-starttime=\"00:05:11\">00:05:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:33\">00:05:33</a>). Deepti presents an architectural overview showing how Elastic gathers context from multiple systems, feeds it to agents, and dispatches it to endpoints such as Microsoft Foundry, Copilot Studio, or other supported environments.</p>\n\n<p><strong>Building Blocks: Context Engine, Skills, and Tools:</strong> Deepti then breaks down the key components of Elastic’s context architecture—agents, skills, connectors, plugins, and the core context engine (<a class=\"medius-aisummary\" data-starttime=\"00:06:38\">00:06:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:20\">00:07:20</a>). The context engine aligns and standardizes semantic meaning across data sources, creating metadata mappings that prevent repetitive computation. She explains this with an example where customer details like name, email, and phone number are recognized as “personal information,” enabling secure and efficient processing (<a class=\"medius-aisummary\" data-starttime=\"00:07:39\">00:07:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:57\">00:07:57</a>). The system’s design improves both security and speed, reducing token usage by up to 30%. She then introduces tools and skills creation—demonstrating how developers can use Elastic SQL, index workflows, and connectors like SharePoint or Slack to define contextual interactions. These tools serve as domain-specific “soldiers” executing precise functions within agents.</p>\n\n<p><strong>Live Demonstration and Use Cases:</strong> In a live demo segment (<a class=\"medius-aisummary\" data-starttime=\"00:09:46\">00:09:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:28\">00:16:28</a>), Deepti shows how a customer support team could use Elastic Agent Builder to unify internal case data with external systems such as GitHub or SharePoint. She queries the system conversationally to retrieve customer interaction summaries, automatically generate dashboards, and integrate enterprise reasoning transparency. Then she demonstrates creating a custom skill for generating a “financial exposure report” and using indexed data tools to perform contextual operations. Finally, she showcases connectors enabling federated search across Slack, SharePoint, and GitHub, including scenarios where an AI agent retrieves a dispute report, finds related policies, checks engineering dependencies, and shares the summarized insights directly into Slack. Deepti ends by reinforcing that Agent Builder drastically minimizes context fragmentation and enables fast, secure, AI-driven business decisions in Azure-integrated environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:00 - 21:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/74755fa7-4099-43f3-98b4-01cb5032629e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/74755fa7-4099-43f3-98b4-01cb5032629e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/74755fa7-4099-43f3-98b4-01cb5032629e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/74755fa7-4099-43f3-98b4-01cb5032629e","speakerIds":["1768245294609build26-1778000884620001iiG0","1768245294609build26-1749604195582001rtwF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Elastic"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/74755fa7-4099-43f3-98b4-01cb5032629e","ogTitle":"Teach AI agents to manage context with Elastic Agent Builder","ogDescription":"Solve AI context limits by enabling agents to manage their own memory. Learn how to prevent bloated prompts and context drift during long tasks, reduce input token usage while ensuring enterprise data governance and scalability. Walk away with next steps for deploying the dynamically loaded skills, conversation context store, selective compaction, and secure external data connectors from Elasticsearch 9.4's Agent Builder  in the Microsoft ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778000884620001iiG0","RegistrantKey":"1768245294609build26-1778000884620001iiG0"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1749604195582001rtwF","RegistrantKey":"1768245294609build26-1749604195582001rtwF"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/Elastic-%20Agent%20Builder%20for%20Microsoft%20Build%20%283%29_1779803490842001RiTQ_1780512582084001bQoe.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"438664a8-5bb9-4f20-9393-dad97c4a8c38","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roelant Dieben, Stephen Simon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774042291880001nzoQ","langLocale":"en-US","sessionId":"1774042291880001nzoQ","sessionInstanceId":"1774042291880001nzoQ","sessionCode":"DEM376","sessionCodeNormalized":"EN-US-DEM376","title":"The story of the Global AI Community that spans every continent","sortTitle":"the story of the global ai community that spans every continent","description":"What starts with a few people passionate about AI becomes a worldwide movement. Global AI Community grew from nothing into thousands of members, hundreds of events, and chapter leads on every continent, all volunteer-run, all free, all real.\n\nIn this session, hear directly from chapter leads about why they joined, what they built, and what community means to them. Plus: how to find your local chapter, start your own, and become part of something bigger than your time zone.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video begins with Ullans, a cloud architect and Global AI Community Board member, welcoming the audience and introducing Simon, the Global AI Community Manager based in India <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>. They clarify that the session is not a demo or product launch, but rather a discussion about the origins and organization of the Global AI Community, its growth, and ways to participate. The opening emphasizes how the community functions globally through its chapters, nurtured by passionate volunteers and local leaders who drive engagement and foster learning through AI initiatives <a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>.</p>\n\n<p><strong>Community Foundation and Growth:</strong> Simon introduces the leadership team, highlighting Hank Goldman as the founder and mentioning other board members, including Sami and Ullans <a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>. They trace the community’s roots to 2017-2018, when AI was emerging but generative AI had yet to bloom. Back then, the movement began with localized meetups and boot camps that eventually expanded despite challenges like COVID-19 <a class=\"medius-aisummary\" data-starttime=\"00:01:23\">00:01:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:17\">00:02:17</a>. The presenters proudly announce milestones: over 200,000 registered members, 200 global chapters, and more than 4,600 events conducted to date <a class=\"medius-aisummary\" data-starttime=\"00:02:36\">00:02:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:56\">00:02:56</a>. This success, they emphasize, has come organically from a “community up” model rather than a brand-driven campaign, focusing on empowering local leaders and connecting AI practitioners at every skill level.</p>\n\n<p><strong>Local Chapters and Event Organization:</strong> Ullans and Simon outline the importance of local chapters, noting their autonomy and deep understanding of regional needs <a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:57\">00:03:57</a>. Chapters are city-based groups led by at least two co-organizers who manage logistics, speakers, venues, and sponsors. Every chapter is encouraged to host one in-person event quarterly, although monthly meetups are even better for fostering engagement <a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>. They acknowledge standout efforts like the Berlin chapter and describe flagship events such as “Agent Camps” (formerly Global AI Boot Camps), where participants engage in hands-on AI workshops <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>. Other key formats include “Agent Cons” (large conferences with international speakers) and “Global AI Nights” (smaller, social learning meetups) — all designed to foster a vibrant global ecosystem.</p>\n\n<p><strong>Participation and Support Structure:</strong> The speakers explain multiple ways to become part of the Global AI Community <a class=\"medius-aisummary\" data-starttime=\"00:10:10\">00:10:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:40\">00:11:40</a>. Interested individuals can join existing chapters, attend scheduled events, or start a new local chapter by completing a simple onboarding process. Chapters receive support through branding materials, templates, newsletters, and access to a YouTube channel with nearly 50,000 subscribers <a class=\"medius-aisummary\" data-starttime=\"00:09:37\">00:09:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:16\">00:10:16</a>. The team underlines the community’s credibility, noting its nine-year history and ongoing expansion. They detail the steps to launch a chapter—finding co-organizers, connecting via Discord, getting onboarding content, and planning hybrid or in-person events—and stress that collaboration with existing chapters ensures sustainable growth <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:40\">00:13:40</a>.</p>\n\n<p><strong>Future Vision and Collaboration:</strong> As AI evolves rapidly, the presenters call for more regional leaders and chapters <a class=\"medius-aisummary\" data-starttime=\"00:13:50\">00:13:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:14\">00:15:14</a>. They describe collaborative opportunities, such as chapters in different countries sharing speakers or resources, and emphasize that engagement through action — organizing, teaching, and building together — brings the most value. Ullans and Simon urge attendees to encourage others to start chapters or join local events, sharing how the community continuously gives back through connections, mentorship, and support in navigating fast-moving AI developments <a class=\"medius-aisummary\" data-starttime=\"00:15:16\">00:15:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a>. Their message highlights empowerment and inclusivity, inviting participants to contribute irrespective of their technology stack — whether Microsoft, AWS, or Google — as Global AI is open to all enthusiasts worldwide <a class=\"medius-aisummary\" data-starttime=\"00:16:40\">00:16:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:58\">00:17:58</a>.</p>\n\n<p><strong>Closing and Key Takeaways:</strong> The session concludes with practical details and encouragement. The speakers mention ongoing opportunities to meet them at the experts’ booth and a table talk session scheduled for the following morning <a class=\"medius-aisummary\" data-starttime=\"00:17:37\">00:17:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:57\">00:17:57</a>. They also clarify the distinctions between the agent camp (small, hands-on boot camps) and agent cons (large conferences with international reach), citing recent examples like the Mountain View event that drew 400 participants from multiple tech communities <a class=\"medius-aisummary\" data-starttime=\"00:19:24\">00:19:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:05\">00:20:05</a>. The presenters close by inviting questions, thanking attendees, and reinforcing their vision: a global, collaborative AI community that thrives on learning, sharing, and helping one another grow <a class=\"medius-aisummary\" data-starttime=\"00:20:49\">00:20:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:53\">00:20:53</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:10 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:35:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9b441553-0170-41c9-9cc5-34040734f410","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9b441553-0170-41c9-9cc5-34040734f410","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9b441553-0170-41c9-9cc5-34040734f410","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9b441553-0170-41c9-9cc5-34040734f410","speakerIds":["1768245294609build26-1770888477835001mAZA","1768245294609build26-1774323608465001D4tS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.859Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9b441553-0170-41c9-9cc5-34040734f410","ogTitle":"The story of the Global AI Community that spans every continent","ogDescription":"What starts with a few people passionate about AI becomes a worldwide movement. Global AI Community grew from nothing into thousands of members, hundreds of events, and chapter leads on every continent, all volunteer-run, all free, all real.\n\nIn this session, hear directly from chapter leads about why they joined, what they built, and what community means to them. Plus: how to find your local chapter, start your own, and become part of something bigger than your time zone.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770888477835001mAZA","RegistrantKey":"1768245294609build26-1770888477835001mAZA"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774323608465001D4tS","RegistrantKey":"1768245294609build26-1774323608465001D4tS"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"d62f1864-8a39-4bb4-87a3-31dd49e1baf9","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chu Lahlou","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018701907001r2xQ","langLocale":"en-US","sessionId":"1774018701907001r2xQ","sessionInstanceId":"1774018701907001r2xQ","sessionCode":"DEM331","sessionCodeNormalized":"EN-US-DEM331","title":"Turn APIs, tools, and data into real agent velocity","sortTitle":"turn apis, tools, and data into real agent velocity","description":"Most teams already have the APIs and data they need. In this demo, see how Foundry enables agents to securely call real APIs, use tools, and act on enterprise data. Learn how Foundry IQ provides grounding and context, while runtime governance ensures tool usage is observable and controlled—turning existing systems into callable agent capabilities.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Problem Context:</strong> The video begins by explaining how AI agents struggle when exposed to unstructured, real-world content such as scanned PDFs, lengthy emails, and office documents containing tables, figures, or multimedia (<a class=\"medius-aisummary\" data-starttime=\"00:01:08\">00:01:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>). While large language models can reason well, they cannot truly “read” or interpret such data effectively, leading to poor reliability and higher costs when raw content is processed directly. To address these challenges, the concept of “content understanding” is introduced as a method to transform messy, multimodal inputs into clean, structured outputs (<a class=\"medius-aisummary\" data-starttime=\"00:01:47\">00:01:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:13\">00:02:13</a>), formatted as Markdown or JSON containing agent-ready inputs.</p>\n\n<p><strong>Scenario Demonstration and Setup:</strong> A sample use case illustrates how content understanding can streamline workflows. In this example, an engineer receives a 6:47 AM alert about signal degradation affecting 42 customers (<a class=\"medius-aisummary\" data-starttime=\"00:02:22\">00:02:22</a>). Normally, the engineer must manually search through documents and media files, but the demo shows how content understanding automates parsing and structuring these materials (<a class=\"medius-aisummary\" data-starttime=\"00:02:46\">00:02:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:13\">00:03:13</a>). It employs the Azure Foundry endpoint and SDK to initialize a client and make prebuilt analyzer calls like “document search,” which captures layout, tables, barcodes, figures, and selection marks (<a class=\"medius-aisummary\" data-starttime=\"00:03:17\">00:03:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:28\">00:04:28</a>).</p>\n\n<p><strong>Document Analysis and Comparison:</strong> The demo continues with analysis of a site maintenance log using the document search analyzer (<a class=\"medius-aisummary\" data-starttime=\"00:04:47\">00:04:47</a>). Within seconds, rich structured output is generated preserving tables, decoding QR codes, and capturing checkmarks. A comparison is made with a local PDF parser which fails to maintain structure or detect key visual elements (<a class=\"medius-aisummary\" data-starttime=\"00:05:37\">00:05:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a>). The video explains that content understanding turns this data into agent-readable formats for reliable reasoning, significantly improving quality and reducing guesswork for LLMs.</p>\n\n<p><strong>Multimodal Analysis and Customization:</strong> Later sections demonstrate handling of audio and video files using prebuilt multimodal analyzers (<a class=\"medius-aisummary\" data-starttime=\"00:06:25\">00:06:25</a>). Audio examples are parsed into timestamped dialogue summaries, while videos yield context-aware insights about fiber strike repairs (<a class=\"medius-aisummary\" data-starttime=\"00:09:25\">00:09:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:33\">00:09:33</a>). Custom analyzers allow users to design bespoke extraction schemas focused on specific business fields. Classification and routing streamline processing by sending unknown document types to appropriate analyzers automatically (<a class=\"medius-aisummary\" data-starttime=\"00:11:06\">00:11:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:09\">00:13:09</a>), ensuring structured results without manual intervention.</p>\n\n<p><strong>Agent Reasoning and Integration:</strong> The next act demonstrates four-step agent reasoning using structured inputs from content understanding (<a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>). The agent diagnoses root causes—such as mechanical failure at Vault TV3—then plans materials within an approved budget and dispatches instructions to relevant personnel (<a class=\"medius-aisummary\" data-starttime=\"00:14:43\">00:14:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:32\">00:15:32</a>). Integration with the Microsoft Agent Framework is then shown, where a context provider hooks into agent loops to automatically process attachments, retrieve structured data, and deliver it for reasoning without additional manual setup (<a class=\"medius-aisummary\" data-starttime=\"00:16:02\">00:16:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:02\">00:18:02</a>).</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> The session concludes by comparing SDK and Agent Framework implementations, noting their shared underpinning service with different integration shapes (<a class=\"medius-aisummary\" data-starttime=\"00:18:07\">00:18:07</a>). Recommended practices include starting with prebuilt analyzers for unknown document types, using classifiers for automatic routing, and applying the \"To-LLM input helper\" function to format outputs and reduce token counts (<a class=\"medius-aisummary\" data-starttime=\"00:19:55\">00:19:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:20:24\">00:20:24</a>). The video closes by inviting viewers to explore full demo code on GitHub and attend a related breakout session on upcoming advancements like Generative Tech Mode and Foundry integration (<a class=\"medius-aisummary\" data-starttime=\"00:21:19\">00:21:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:47\">00:21:47</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"16:30 - 16:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:55:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/f08b2f52-65fb-4621-90fd-7813993f93ca","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/f08b2f52-65fb-4621-90fd-7813993f93ca","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/f08b2f52-65fb-4621-90fd-7813993f93ca","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/f08b2f52-65fb-4621-90fd-7813993f93ca","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/f08b2f52-65fb-4621-90fd-7813993f93ca","speakerIds":["1768245294609build26-1775581764225001HWz2"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.457Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/f08b2f52-65fb-4621-90fd-7813993f93ca","ogTitle":"Turn APIs, tools, and data into real agent velocity","ogDescription":"Most teams already have the APIs and data they need. In this demo, see how Foundry enables agents to securely call real APIs, use tools, and act on enterprise data. Learn how Foundry IQ provides grounding and context, while runtime governance ensures tool usage is observable and controlled—turning existing systems into callable agent capabilities.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775581764225001HWz2","RegistrantKey":"1768245294609build26-1775581764225001HWz2"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/DEM331","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Grounding","logicalValue":"Grounding"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"fa972117-5161-4e50-acdf-bbfb5fa83777","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dave Citron","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776889045397001i1x5","langLocale":"en-US","sessionId":"1776889045397001i1x5","sessionInstanceId":"1776889045397001i1x5","sessionCode":"DEM323","sessionCodeNormalized":"EN-US-DEM323","title":"Under the hood of Microsoft AI models","sortTitle":"under the hood of microsoft ai models","description":"Microsoft AI (MAI) just announced a family of new models including new Thinking, Coding, Voice, Transcription, and Image models. \u00A0In this session, a leader from MAI will share an open and scientific walk through of what it takes to train our models, what we learn along the way, and how those learnings are designed into the model architectures, features, and capabilities.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The session opens at <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a> with host Dave Citron, CVP of Products at Microsoft AI, welcoming attendees and explaining that this is an \"under the hood\" deep dive into models announced earlier that day. He describes the agenda at <a class=\"medius-aisummary\" data-starttime=\"00:00:32\">00:00:32</a>—starting with philosophy, followed by a walkthrough of seven released models, a deep exploration of the \"Thinking 1\" reasoning model, and closing with a detailed look at Microsoft’s \"Frontier Tuning\" methodology. This introduction sets the expectation that participants will learn precisely how these new AI models were constructed, trained, and optimized for enterprise-grade performance and customization.</p>\n\n<p><strong>AI Model Portfolio Overview:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:01:04\">00:01:04</a>, Citron highlights the seven core models announced that morning, spanning image generation, transcription, voice, coding, and reasoning. \"Image 2.5\" and its \"Flash\" variant deliver high-quality image generation at reduced cost. \"Transcribe 1.5\" offers record-setting accuracy across 43 languages at five times the performance speed of competitors. \"Voice 2\" provides rich, natural speech generation with emotional control and ultra-low latency, demonstrated at <a class=\"medius-aisummary\" data-starttime=\"00:06:20\">00:06:20</a>. \"Code 1 Flash\" enables efficient, high-speed code generation already integrated into GitHub and Visual Studio Code. Finally, Citron introduces \"Thinking 1,\" Microsoft’s first frontier reasoning model performing at 97% accuracy on core benchmarks, built entirely through their own training loops without distillation.</p>\n\n<p><strong>Philosophy and Design Principles:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:02:29\">00:02:29</a> to <a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>, Citron emphasizes Microsoft’s core development philosophy—dubbed \"Humanist Superintelligence.\" This approach is defined by a commitment to AI that augments rather than replaces human potential. Three driving tenets—human-first design, service-oriented architecture, and developer empowerment—shape their modeling decisions. Notably, Microsoft avoids distillation, preferring to \"hill climb from scratch\" so that each model earns its own capabilities through authentic data learning loops. This methodology ensures full transparency and control, with clean datasets, no inherited third-party weights, and complete auditability. The foundation supports reproducible, safe, and ethically aligned AI development with traceable lineage for enterprise deployment.</p>\n\n<p><strong>Model Deep Dive – Architecture and Reinforcement Learning:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>, Citron focuses on \"Thinking 1\" and outlines its architectural composition—a 35-billion active parameter mixture-of-experts framework capable of handling a 256k context window. The model was trained on over 30 trillion in-house tokens, intentionally excluding synthetic or AI-generated data. At <a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>, he explains the reinforcement learning (RL) process, powered by a GRPO algorithm that iteratively scores and reinforces valid reasoning chains. RL was stabilized through five innovations allowing reliable capability growth over thousands of training steps, resulting in strong benchmark performances. Safety is handled through a dedicated RL climb using human preference data and red teaming with over 2,100 adversarial scenarios, ensuring high safety and helpfulness. Benchmarks presented at <a class=\"medius-aisummary\" data-starttime=\"00:13:16\">00:13:16</a> show the model’s excellence across reasoning, coding, and scientific domains, proving that capabilities were “learned, not inherited.”</p>\n\n<p><strong>Frontier Tuning and Customization:</strong> In the segment beginning at <a class=\"medius-aisummary\" data-starttime=\"00:14:27\">00:14:27</a>, Citron introduces Microsoft’s \"Frontier Tuning,\" which empowers organizations to run their own private modeling and optimization cycles atop the foundation models. The tuning system allows enterprises to train within secure environments using their proprietary data, achieving personalized AI behavior tailored to context, workflows, and domain expertise. The process emphasizes privacy, efficiency, control, and continuous improvement, transforming usage feedback into progressive model evolution. A case study with Land O’Lakes at <a class=\"medius-aisummary\" data-starttime=\"00:16:04\">00:16:04</a> demonstrates this advantage—Frontier-tuned Thinking 1 Flash achieved 89.3% quality scores and 10X cost efficiency while outperforming generalist models on production tasks.</p>\n\n<p><strong>Conclusion and Availability:</strong> In closing at <a class=\"medius-aisummary\" data-starttime=\"00:16:48\">00:16:48</a>, Citron summarizes that Image 2.5, Voice 2, and Transcribe 1.5 are live today on Foundry and Microsoft’s Mai Playground, accessible even through mobile devices, while Thinking 1 is entering private preview through Foundry. \"Code 1 Flash\" is available in Visual Studio Code. All models will also deploy across Base10, OpenRouter, and Fireworks platforms. He invites users to visit Foundry or Microsoft.ai for detailed reports, model cards, pricing, and documentation. The talk ends at <a class=\"medius-aisummary\" data-starttime=\"00:17:49\">00:17:49</a> with encouragement to test demos, share feedback, and follow Microsoft AI’s forthcoming innovations and research as the lab continues to expand its frontier development roadmap.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/737e6687-341c-48f1-955a-62f5f84e88a1","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/737e6687-341c-48f1-955a-62f5f84e88a1","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/737e6687-341c-48f1-955a-62f5f84e88a1","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/737e6687-341c-48f1-955a-62f5f84e88a1","speakerIds":["1768245294609build26-1777069301321001b1id"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.387Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/737e6687-341c-48f1-955a-62f5f84e88a1","ogTitle":"Under the hood of Microsoft AI models","ogDescription":"Microsoft AI (MAI) just announced a family of new models including new Thinking, Coding, Voice, Transcription, and Image models. \u00A0In this session, a leader from MAI will share an open and scientific walk through of what it takes to train our models, what we learn along the way, and how those learnings are designed into the model architectures, features, and capabilities.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Festival Pavilion, Theater A","logicalValue":"Festival Pavilion, Theater A"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) In-person","SpeakerId":"1768245294609build26-1777069301321001b1id","RegistrantKey":"1768245294609build26-1777069301321001b1id"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"6fdbcab5-eafd-491c-8aab-44990da413dc","startDateTime":"2026-06-03T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jim Bennett","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776289863408001ZN64","langLocale":"en-US","sessionId":"1776289863408001ZN64","sessionInstanceId":"1776289863408001ZN64","sessionCode":"DEM361","sessionCodeNormalized":"EN-US-DEM361","title":"Understand and fix Agent Framework apps with observability and evals","sortTitle":"understand and fix agent framework apps with observability and evals","description":"Your AI apps are getting more complex, with multiple agents, tools, and different orchestration patterns. This makes them harder to understand, debug, and test. This session shows you how to visualize the decisions your LLMs are making in complex Microsoft Agent Framework applications, using open standards and open source tooling to provide you with instrumentation and observability. You'll also see how you can use another LLM as a judge to evaluate how well your AI app is working.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The video begins with <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a> Jim Bennett, a Microsoft MVP, introducing himself and explaining that he will discuss how developers can understand what their AI agents are doing under the hood. He starts by asking the audience who has heard of AI and who is actively building agents (<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:33\">00:00:33</a>), revealing that few people actually know whether their agents work correctly or what they do internally. This sets the stage for a hands-on session focused on demystifying agent frameworks and adding observability.</p>\n\n<p><strong>Understanding AI Agents and Frameworks:</strong> Jim demonstrates a simple agent (<a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a>), describing how it’s composed of an LLM, instructions, and tools such as vendor and policy lookup utilities. He explains that his agent runs on Microsoft Foundry and is designed as a procurement evaluator that automatically approves or denies purchase requests based on internal data. The agent’s code is minimal, but the decision-making happens inside a black box—developers simply run “agent.run” and hope it behaves correctly. Jim emphasizes that users have no visibility into which tools are triggered, in what order, or how decisions are made.</p>\n\n<p><strong>Adding Observability with Open Telemetry and Open Inference:</strong> Moving into improvement strategies (<a class=\"medius-aisummary\" data-starttime=\"00:03:25\">00:03:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:12\">00:05:12</a>), Jim introduces the concept of observability through Open Telemetry, a framework commonly used in web applications to track performance and errors. He explains how Open Inference extends Open Telemetry for AI, allowing developers to collect and visualize semantic telemetry data from LLMs. Using the Phoenix open source back end (<a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>), he demonstrates a workflow that exposes detailed traces of agent operations — showing tool calls, token counts, inputs, and model responses. This makes it possible to understand every decision the framework makes and enables developers to strip sensitive data or manage telemetry pipelines using standard industry processors.</p>\n\n<p><strong>Testing with Evals and AI Judges:</strong> Jim transitions to the question of reliability and testing (<a class=\"medius-aisummary\" data-starttime=\"00:07:27\">00:07:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:14\">00:13:14</a>). He argues that deterministic unit tests don’t work for LLMs due to non-deterministic outputs. Instead, developers can use “evals,” which are AI-driven audits that assess whether the model’s decisions are grounded in available data. He illustrates a prompt-driven evaluation where the LLM acts as a judge, reviewing agent outputs for approval rationale grounded in evidence from tools. Through Phoenix, he shows how evals produce simple deterministic scores (supported/unsupported) and reveals that his agent is only correct around half the time. Using these results, developers can pinpoint logic or model flaws, enabling automated testing cycles similar to standard test-driven development workflows.</p>\n\n<p><strong>Model Selection and Self-Improving Agents:</strong> Jim then demonstrates how telemetry and evals combine for self-improvement (<a class=\"medius-aisummary\" data-starttime=\"00:14:05\">00:14:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:29\">00:16:29</a>). He compares results from GPT-4.1 and GPT-5.4, showing improved effectiveness from 50% to 90% simply by changing the model. He explains how this data-driven testing can feed into self-improving loops using coding agents such as GitHub Copilot or Claude. These agents can analyze telemetry data, run evaluations, tweak prompts, alter tool descriptions, and iterate several cycles automatically to reach optimal performance. Jim cautions that AI agents, like humans, will never achieve 100% accuracy, but developers can aim for stable models performing near 90% effectiveness using these improvement loops.</p>\n\n<p><strong>Conclusion and Resources:</strong> In closing (<a class=\"medius-aisummary\" data-starttime=\"00:16:36\">00:16:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:23\">00:18:23</a>), Jim recaps the session: agents are inherently black boxes, and adding observability through Open Inference reveals their internal workings. Using evals enables systematic AI testing, laying the foundation for self-improving systems. He encourages the audience to experiment with Phoenix, which is free and open source, and provides QR codes for downloading the code and connecting with him on LinkedIn. Despite his Azure credits expiring, Jim assures participants they can run everything locally and invites further questions to continue exploring agent monitoring and evaluation techniques.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"22:30 - 22:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/65ab30d5-8261-422e-8be7-119ad843b8ee","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/65ab30d5-8261-422e-8be7-119ad843b8ee","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/65ab30d5-8261-422e-8be7-119ad843b8ee","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/65ab30d5-8261-422e-8be7-119ad843b8ee","speakerIds":["1768245294609build26-1776289931354001kW0l"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.903Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/65ab30d5-8261-422e-8be7-119ad843b8ee","ogTitle":"Understand and fix Agent Framework apps with observability and evals","ogDescription":"Your AI apps are getting more complex, with multiple agents, tools, and different orchestration patterns. This makes them harder to understand, debug, and test. This session shows you how to visualize the decisions your LLMs are making in complex Microsoft Agent Framework applications, using open standards and open source tooling to provide you with instrumentation and observability. You'll also see how you can use another LLM as a judge to evaluate how well your AI app is working.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776289931354001kW0l","RegistrantKey":"1768245294609build26-1776289931354001kW0l"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"0541280f-d88f-4718-b32e-3fe99bb8259f","startDateTime":"2026-06-02T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dustin Campbell, Mads Torgersen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018710414001rtag","langLocale":"en-US","sessionId":"1774018710414001rtag","sessionInstanceId":"1774018710414001rtag","sessionCode":"DEM304","sessionCodeNormalized":"EN-US-DEM304","title":"Union types in C#","sortTitle":"union types in c#","description":"Union types are coming to C#! Unions model closed sets of data shapes, as commonly seen in e.g. wire protocols. Mads and Dustin explore the clean expression of intent and the confidence and elegance that unions lend to consuming code.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Context:</strong> The session opens with Matt Torgersen and Dustin Campbell from the C# team introducing themselves and announcing that they will discuss a single new feature in C# 15 called Union Types <a class=\"medius-aisummary\" data-starttime=\"00:00:23\">00:00:23</a>. They note that while features like this exist in other languages such as TypeScript and functional languages, the C# team has designed their own \"Goldilocks\" version—a balanced and practical approach to union types. They mention that C# 15 is releasing in November and that the functionality is already available in preview <a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>.</p>\n\n<p><strong>Motivation and Basic Example:</strong> Transitioning into a demo <a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>, the presenters show code modeling animals and pets. Previously, developers would have to treat distinct classes like Dog and Cat as generic objects, losing type safety. For example, a Shark could be mistakenly treated as a Pet, something the compiler could not prevent. With union types, they now define a \"union Pet = Dog | Cat | Bird\" <a class=\"medius-aisummary\" data-starttime=\"00:03:28\">00:03:28</a>. This enables the compiler to validate that only those specified types can be pets, offering exhaustiveness checking in switch expressions and preventing unwanted conversions. The demonstration also highlights how implicit conversions and pattern matching work seamlessly with union types <a class=\"medius-aisummary\" data-starttime=\"00:05:02\">00:05:02</a>.</p>\n\n<p><strong>Members and Extended Usage:</strong> The team then shows that unions can contain function members such as computed properties <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>. For example, a Pet union can include a property like Description that uses pattern matching to derive textual summaries for each pet type. However, unions are intended to represent the state of their underlying member types and should not introduce additional mutable state. This design encourages treating unions as lightweight, composable abstractions rather than containers of mutable behavior. The presenters emphasize that beyond pets and animals, the same concept can apply to any domain, aligning with real-world software design challenges <a class=\"medius-aisummary\" data-starttime=\"00:08:13\">00:08:13</a>.</p>\n\n<p><strong>Practical Scenarios and Generics:</strong> A second demo presents unions in a more realistic programming scenario involving network results <a class=\"medius-aisummary\" data-starttime=\"00:08:46\">00:08:46</a>. They define a generic Result<T> union composed of two case types: Success<T> and Error. This pattern allows functions to return either a successful result or an error state in a strongly typed way, improving code reliability. When looping over a queue of result values, the switch expression enforces that both cases—success and error—are handled, preventing missing-case bugs. The example demonstrates that unions can be generic, support deconstruction, and provide the same exhaustiveness and implicit conversion guarantees as simpler cases <a class=\"medius-aisummary\" data-starttime=\"00:12:04\">00:12:04</a>.</p>\n\n<p><strong>Implementation Details and Customization:</strong> Diving deeper under the hood, Matt and Dustin explain that unions in C# are implemented as structs with an object field storing the underlying value <a class=\"medius-aisummary\" data-starttime=\"00:14:18\">00:14:18</a>. The feature is built using new attributes such as [Union], and developers can technically define their own unions manually using constructors and the IUnion interface <a class=\"medius-aisummary\" data-starttime=\"00:15:06\">00:15:06</a>. They outline two main reasons one might customize implementation: achieving higher performance by avoiding boxing allocations, or adapting legacy types that developers previously built to simulate unions. Through these mechanisms, developers gain granular control over type modeling and performance behavior <a class=\"medius-aisummary\" data-starttime=\"00:18:12\">00:18:12</a>.</p>\n\n<p><strong>Design Philosophy, Q&A, and Conclusion:</strong> In the concluding portion, the team compares their approach to TypeScript’s structural unions and explains why C# uses nominal, declared unions instead <a class=\"medius-aisummary\" data-starttime=\"00:22:26\">00:22:26</a>. Because .NET requires runtime type representation, structural unions like in TypeScript are impractical. Instead, C# unions exist both at compile-time and runtime, offering strong typing and runtime inspection capabilities. After nearly five years of design iterations <a class=\"medius-aisummary\" data-starttime=\"00:24:27\">00:24:27</a>, the team settled on this balanced model that integrates smoothly with C# language semantics. They end by noting that unions are supported in the .NET 11 and Visual Studio preview releases and invite developers to explore the feature and provide feedback in the expert zone after the session <a class=\"medius-aisummary\" data-starttime=\"00:25:00\">00:25:00</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"23:00 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","speakerIds":["1768245294609build26-1774284998022001prSB","1768245294609build26-1774284998469001pFjY"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.79Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/5823d341-1ee8-4b7a-8fdd-2abf8469ddd4","ogTitle":"Union types in C#","ogDescription":"Union types are coming to C#! Unions model closed sets of data shapes, as commonly seen in e.g. wire protocols. Mads and Dustin explore the clean expression of intent and the confidence and elegance that unions lend to consuming code.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998022001prSB","RegistrantKey":"1768245294609build26-1774284998022001prSB"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998469001pFjY","RegistrantKey":"1768245294609build26-1774284998469001pFjY"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build26/csharp/union-types","backpackLink":null,"label":"Try out C# unions in preview","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/DEM304","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"1fe0dc14-f87a-4ad7-b0ad-bd1567a4bab0","startDateTime":"2026-06-02T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Craig Loewen, Pooja Trivedi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018719239001rqgK","langLocale":"en-US","sessionId":"1774018719239001rqgK","sessionInstanceId":"1774018719239001rqgK","sessionCode":"DEM346","sessionCodeNormalized":"EN-US-DEM346","title":"WSL improvements and the new Containers CLI and APIs","sortTitle":"wsl improvements and the new containers cli and apis","description":"Learn about the new core end to end improvements to Windows Subsystem for Linux and the new Container CLI and Apis for your applications\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"<p><strong>Introduction and Key Announcements:</strong> The session begins with Pooja Trivedi and Craig introducing themselves and unveiling a new capability known as WSL Containers (<a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>). They highlight three central benefits: the ability to run Linux containers natively on Windows, full integration within the Windows Subsystem for Linux (WSL), and enterprise readiness through compatibility with management tools such as Microsoft Defender for Endpoint and Intune (<a class=\"medius-aisummary\" data-starttime=\"00:00:56\">00:00:56</a>). Pooja and Craig emphasize that no extra installation is required — the feature will come bundled as part of WSL updates — ensuring a secure and seamless experience for both developers and enterprise IT administrators alike. The presenters then prepare to transition from an overview into a practical demonstration showcasing this functionality via the command line.</p>\n\n<p><strong>CLI Demonstration and Basic Container Operations:</strong> Pooja demonstrates how WSL Containers can be accessed and managed using a new tool called WSLC, available after a simple WSL update (<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>). She shows that the WSLC command structure mirrors familiar Linux container tooling, allowing users to run commands like “container run,” “container list,” and “container attach” with Linux-like syntax (<a class=\"medius-aisummary\" data-starttime=\"00:02:06\">00:02:06</a>). Through the example of launching and attaching to a Debian-based container, Pooja demonstrates interactive shells, detaching and reattaching, listing running containers, and terminating sessions efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:03:35\">00:03:35</a>). This demo affirms that developers can now perform full local Linux container operations on Windows with minimal setup and familiar CLI semantics.</p>\n\n<p><strong>Building Custom Images and GPU-Accelerated AI Workloads:</strong> The demonstration continues with a deeper look at container image creation using WSLC (<a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>). Pooja introduces a sample container file that bundles dependencies, source code, and runtime instructions to create custom environments. She walks through building an example Python-based microservice and mapping ports between the Linux container and the Windows host for seamless testing (<a class=\"medius-aisummary\" data-starttime=\"00:05:47\">00:05:47</a>). Next, she shows an AI workflow where a Jupyter Notebook container utilizes the host GPU to run deep learning models efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:07:07\">00:07:07</a>). Using the \"--gpus all\" option, she demonstrates fine-tuning a model and compiling kernels with PyTorch, underscoring that developers can now leverage the full performance of Windows hardware for AI and data workloads inside Linux containers. The combination of GPU acceleration, code portability, and interactive tools bridges the gap between Windows development and Linux computational environments.</p>\n\n<p><strong>API Integration, App Developer Use Cases, and Partner Example:</strong> Craig returns to explain how WSL Containers can be integrated into applications not only through the CLI but also through a new NuGet-based API (<a class=\"medius-aisummary\" data-starttime=\"00:09:07\">00:09:07</a>). He outlines the audience for this feature — container developers, independent software vendors, and IT administrators — describing Microsoft’s goal to unify development tooling and enterprise control. The team highlights improvements in virtualization and cross-OS file performance, benefiting all container technologies including Docker and Rancher Desktop. As a concrete example, Craig runs a demo using the Linux-based Moonray rendering engine (<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>). By invoking Moonray directly via a Windows executable, the system transparently manages a Linux container in the background, rendering complex graphics without user intervention. This shows the API’s ability to run Linux-native workloads as seamless Windows apps, even for high-performance graphical computing.</p>\n\n<p><strong>Developer Workflow Demonstrations and Container Safety:</strong> To illustrate development integration, the presenters show a Visual Studio project embedding a container build process using the WSL Containers SDK (<a class=\"medius-aisummary\" data-starttime=\"00:14:21\">00:14:21</a>). They demonstrate a playful AI stock-trading bot named Herbert running inside a container, visually observed via port-mounted connections. The scenario humorously includes file manipulation actions to highlight container-to-host interactions and the isolation boundaries ensuring data safety (<a class=\"medius-aisummary\" data-starttime=\"00:16:04\">00:16:04</a>). Even though the container can access certain directories, its scope is constrained by explicitly defined mount points, minimizing risk. This portion reinforces the flexibility and secure design of Windows-integrated container environments that allow interactive, AI-driven or graphical workloads while maintaining OS-level isolation.</p>\n\n<p><strong>Architecture Deep Dive and Conclusion:</strong> Pooja concludes with an architectural breakdown of how WSL Containers operate under the hood (<a class=\"medius-aisummary\" data-starttime=\"00:17:00\">00:17:00</a>). Each application or command-line session runs within an independent lightweight utility VM managed by the WSL service. Communication between Windows and Linux environments happens through hypervisor sockets (HV sockets) for performance and security, while storage is isolated using per-application virtual disks (<a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>). File sharing between host and containers is powered by VirtioFS, which doubles speed compared to older Plan 9 implementations. Before closing, she mentions additional updates including Azure Linux 4.0’s availability and its upcoming inclusion as a WSL distribution (<a class=\"medius-aisummary\" data-starttime=\"00:21:52\">00:21:52</a>). The presenters wrap up by announcing a forthcoming public preview by the end of June, encouraging developers to experiment, provide feedback, and follow ongoing progress through the Microsoft WSL open-source repository (<a class=\"medius-aisummary\" data-starttime=\"00:22:57\">00:22:57</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":5,"TimeSlot":"21:40 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":25,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/4025f713-c725-4c66-a946-2ab72f10c1ad","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4025f713-c725-4c66-a946-2ab72f10c1ad","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4025f713-c725-4c66-a946-2ab72f10c1ad","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4025f713-c725-4c66-a946-2ab72f10c1ad","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4025f713-c725-4c66-a946-2ab72f10c1ad","speakerIds":["1768245294609build26-1758133761538001NuMh","1768245294609build26-1774284998933001poGL"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.857Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4025f713-c725-4c66-a946-2ab72f10c1ad","ogTitle":"WSL improvements and the new Containers CLI and APIs","ogDescription":"Learn about the new core end to end improvements to Windows Subsystem for Linux and the new Container CLI and Apis for your applications\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Demo","logicalValue":"Demo"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"C++","logicalValue":"C++"},{"displayValue":"C","logicalValue":"C"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758133761538001NuMh","RegistrantKey":"1768245294609build26-1758133761538001NuMh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998933001poGL","RegistrantKey":"1768245294609build26-1774284998933001poGL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/DEM346","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Linux","logicalValue":"Linux"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"WSL","logicalValue":"WSL"},{"displayValue":"Linux Containers","logicalValue":"Linux Containers"},{"displayValue":"Distributions","logicalValue":"Distributions"},{"displayValue":"Windows applications","logicalValue":"Windows applications"}]},{"@search.score":1.0,"documentId":"05995ddc-e530-4bf9-8fd2-40dc90159bd9","startDateTime":"2026-06-02T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711414001rPGE","langLocale":"en-US","sessionId":"1774018711414001rPGE","sessionInstanceId":"1774018711414001rPGE","sessionCode":"LTG401","sessionCodeNormalized":"EN-US-LTG401","title":"10 ways to maximize GitHub Copilot","sortTitle":"10 ways to maximize github copilot","description":"It started as autocomplete, but that's so 2024. What can GitHub Copilot really do in 2026? Join us for our 10 favorite hidden gems. Number will shock you!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:30 - 22:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711414001rPGE_20260512022341854.avif","speakerIds":["1768245294609build26-1772472228939001By6I"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.852Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711414001rPGE_20260512022341854.avif","ogTitle":"10 ways to maximize GitHub Copilot","ogDescription":"It started as autocomplete, but that's so 2024. What can GitHub Copilot really do in 2026? Join us for our 10 favorite hidden gems. Number will shock you!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"ca4f01ff-236a-4b8a-8079-9890c42ec997","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aaron Powell","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778262935558001RpbP","langLocale":"en-US","sessionId":"1778262935558001RpbP","sessionInstanceId":"1778262935558001RpbP","sessionCode":"LTG401-R1","sessionCodeNormalized":"EN-US-LTG401-R1","title":"10 ways to maximize GitHub Copilot (Repeat)","sortTitle":"10 ways to maximize github copilot (repeat)","description":"It started as autocomplete, but that's so 2024. What can GitHub Copilot really do in 2026? Join us for our 10 favorite hidden gems. Number will shock you!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778262935558001RpbP_20260512031931458.avif","speakerIds":["1768245294609build26-1772472228939001By6I"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.775Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778262935558001RpbP_20260512031931458.avif","ogTitle":"10 ways to maximize GitHub Copilot (Repeat)","ogDescription":"It started as autocomplete, but that's so 2024. What can GitHub Copilot really do in 2026? Join us for our 10 favorite hidden gems. Number will shock you!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772472228939001By6I","RegistrantKey":"1768245294609build26-1772472228939001By6I"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"31b4bcb5-40e6-48b6-baf3-f58826cf6db1","startDateTime":"2026-06-02T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mohammad Abuomar, Mike Shi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717724001rECb","langLocale":"en-US","sessionId":"1774018717724001rECb","sessionInstanceId":"1774018717724001rECb","sessionCode":"LTG430","sessionCodeNormalized":"EN-US-LTG430","title":"Agent Control Specification: Universal Control Layer for Safe Agents","sortTitle":"agent control specification: universal control layer for safe agents","description":"AI agents are rapidly gaining autonomy—but most still rely on prompts and ad hoc checks for safety. The Agent Control Specification introduces a universal runtime standard that adds enforceable guardrails across frameworks, models, and clouds. This talk shows how portable policy definitions can govern tool use, sensitive actions, approvals, and outputs in real time—turning agent safety from best effort into deterministic control without slowing innovation.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:30 - 22:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717724001rECb_20260603024726288.avif","speakerIds":["1768245294609build26-1777687154991001jRn7","1768245294609build26-1776099961753001AJjB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.856Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717724001rECb_20260603024726288.avif","ogTitle":"Agent Control Specification: Universal Control Layer for Safe Agents","ogDescription":"AI agents are rapidly gaining autonomy—but most still rely on prompts and ad hoc checks for safety. The Agent Control Specification introduces a universal runtime standard that adds enforceable guardrails across frameworks, models, and clouds. This talk shows how portable policy definitions can govern tool use, sensitive actions, approvals, and outputs in real time—turning agent safety from best effort into deterministic control without slowing innovation.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776099961753001AJjB","RegistrantKey":"1768245294609build26-1776099961753001AJjB"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777687154991001jRn7","RegistrantKey":"1768245294609build26-1777687154991001jRn7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Guardrails","logicalValue":"Guardrails"}]},{"@search.score":1.0,"documentId":"0508c9b3-8b04-487d-a50f-b3fa5b8554e9","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Carlo","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293184274001L59w","langLocale":"en-US","sessionId":"1776293184274001L59w","sessionInstanceId":"1776293184274001L59w","sessionCode":"LTG458","sessionCodeNormalized":"EN-US-LTG458","title":"Agentic Apps on Microsoft Fabric: Mind-Blowing Interactive Experiences","sortTitle":"agentic apps on microsoft fabric: mind-blowing interactive experiences","description":"Imagine turning your Microsoft Fabric semantic models into stunning, interactive apps in minutes. Less Clicking More Building. Discover the brand-new agentic development experience that lets you choose the agents to build reporting. No deep coding, no complex frameworks, just point, describe, and watch agents orchestrate everything. Visuals, custom navigation, writeback, you can have it all directly on your Fabric assets. Think no-holds-barred analytics experiences.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:50 - 19:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293184274001L59w_20260512030207009.avif","speakerIds":["1768245294609build26-1776293231988001WPJb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.89Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293184274001L59w_20260512030207009.avif","ogTitle":"Agentic Apps on Microsoft Fabric: Mind-Blowing Interactive Experiences","ogDescription":"Imagine turning your Microsoft Fabric semantic models into stunning, interactive apps in minutes. Less Clicking More Building. Discover the brand-new agentic development experience that lets you choose the agents to build reporting. No deep coding, no complex frameworks, just point, describe, and watch agents orchestrate everything. Visuals, custom navigation, writeback, you can have it all directly on your Fabric assets. Think no-holds-barred analytics experiences.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293231988001WPJb","RegistrantKey":"1768245294609build26-1776293231988001WPJb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"Skills","logicalValue":"Skills"}]},{"@search.score":1.0,"documentId":"1be5af6b-2ec8-421e-af62-6f96d0d582dc","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roelant Dieben","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776292575600001CoIk","langLocale":"en-US","sessionId":"1776292575600001CoIk","sessionInstanceId":"1776292575600001CoIk","sessionCode":"LTG451","sessionCodeNormalized":"EN-US-LTG451","title":"Agentic FinOps: Governing Cost and Quality with Microsoft Foundry","sortTitle":"agentic finops: governing cost and quality with microsoft foundry","description":"Your agents are spending money you can't see. Traditional FinOps tooling was built for workloads humans start and stop, agents don't work that way. In this session, you'll learn how to use Microsoft Foundry's observability plane to get real-time visibility into token consumption, tool call chains, and reasoning traces, and how to turn that data into actual cost governance before your cloud bill becomes a surprise.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:30 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292575600001CoIk_20260512030027885.avif","speakerIds":["1768245294609build26-1770888477835001mAZA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.887Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292575600001CoIk_20260512030027885.avif","ogTitle":"Agentic FinOps: Governing Cost and Quality with Microsoft Foundry","ogDescription":"Your agents are spending money you can't see. Traditional FinOps tooling was built for workloads humans start and stop, agents don't work that way. In this session, you'll learn how to use Microsoft Foundry's observability plane to get real-time visibility into token consumption, tool call chains, and reasoning traces, and how to turn that data into actual cost governance before your cloud bill becomes a surprise.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770888477835001mAZA","RegistrantKey":"1768245294609build26-1770888477835001mAZA"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"14a2eea9-cace-48ea-9e4d-02e42b34cd7f","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"James Codella","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707578001rTn5","langLocale":"en-US","sessionId":"1774018707578001rTn5","sessionInstanceId":"1774018707578001rTn5","sessionCode":"LTG411","sessionCodeNormalized":"EN-US-LTG411","title":"Agentic RAG for fast, low-cost knowledge retrieval on Azure Cosmos DB","sortTitle":"agentic rag for fast, low-cost knowledge retrieval on azure cosmos db","description":"Skip the knowledge graph and ontology construction. The Agentic Retrieval Toolkit delivers high-quality answers to open-ended, multi-hop, and nuanced questions in seconds using GPU-accelerated local models or Microsoft Foundry models and Azure Cosmos DB — combining hybrid full-text and vector search, diversity selection, and sub-question decomposition across large structured and unstructured datasets. You'll leave knowing how to build production-ready knowledge retrieval as long-term agentic memory to boost your agents' accuracy and speed.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:30 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707578001rTn5_20260512021838542.avif","speakerIds":["1768245294609build26-1776347675875001DHR6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707578001rTn5_20260512021838542.avif","ogTitle":"Agentic RAG for fast, low-cost knowledge retrieval on Azure Cosmos DB","ogDescription":"Skip the knowledge graph and ontology construction. The Agentic Retrieval Toolkit delivers high-quality answers to open-ended, multi-hop, and nuanced questions in seconds using GPU-accelerated local models or Microsoft Foundry models and Azure Cosmos DB — combining hybrid full-text and vector search, diversity selection, and sub-question decomposition across large structured and unstructured datasets. You'll leave knowing how to build production-ready knowledge retrieval as long-term agentic memory to boost your agents' accuracy and speed.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[{"displayValue":"CUDA","logicalValue":"CUDA"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347675875001DHR6","RegistrantKey":"1768245294609build26-1776347675875001DHR6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"OFT","logicalValue":"OFT"}]},{"@search.score":1.0,"documentId":"e4a914af-5686-4d9c-8610-6b636fb80d51","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Onur Okutucu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293340144001LRTb","langLocale":"en-US","sessionId":"1776293340144001LRTb","sessionInstanceId":"1776293340144001LRTb","sessionCode":"LTG460","sessionCodeNormalized":"EN-US-LTG460","title":"Agents Gone Wild: The Best (and Worst) AI Agents You Can Build Today","sortTitle":"agents gone wild: the best (and worst) ai agents you can build today","description":"AI agents are everywhere—but not all of them should exist. In this session, we’ll explore how to build powerful, practical AI agents using Microsoft’s modern AI stack, including Azure AI services, Microsoft Foundry, and Copilot-ready integrations. Through real-world examples, we’ll look at useful agents such as a support assistant that summarizes customer tickets, a knowledge agent that answers questions from internal docs, and an automation agent that triages alerts from the cloud infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:00 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293340144001LRTb_20260512030233179.avif","speakerIds":["1768245294609build26-1765878443369001FyEA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.889Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293340144001LRTb_20260512030233179.avif","ogTitle":"Agents Gone Wild: The Best (and Worst) AI Agents You Can Build Today","ogDescription":"AI agents are everywhere—but not all of them should exist. In this session, we’ll explore how to build powerful, practical AI agents using Microsoft’s modern AI stack, including Azure AI services, Microsoft Foundry, and Copilot-ready integrations. Through real-world examples, we’ll look at useful agents such as a support assistant that summarizes customer tickets, a knowledge agent that answers questions from internal docs, and an automation agent that triages alerts from the cloud infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765878443369001FyEA","RegistrantKey":"1768245294609build26-1765878443369001FyEA"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"e5e6cfcf-6180-477e-9638-6eb3a2f9ee42","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Adnan Hendricks","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293411668001DEZ6","langLocale":"en-US","sessionId":"1776293411668001DEZ6","sessionInstanceId":"1776293411668001DEZ6","sessionCode":"LTG461","sessionCodeNormalized":"EN-US-LTG461","title":"AI building using the Foundry Extension for VS Code","sortTitle":"ai building using the foundry extension for vs code","description":"This session does a quick overview on how to create an Ai app using the Foundry Toolkit Extension for VS Code. Attendees will see how to select an Ai model, compare models to select the most appropriate one for your app. Teaching you how to go from zero to hero in 15 minutes.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:50 - 19:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293411668001DEZ6_20260520160655023.avif","speakerIds":["1768245294609build26-1770479415883001tO49"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.888Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293411668001DEZ6_20260520160655023.avif","ogTitle":"AI building using the Foundry Extension for VS Code","ogDescription":"This session does a quick overview on how to create an Ai app using the Foundry Toolkit Extension for VS Code. Attendees will see how to select an Ai model, compare models to select the most appropriate one for your app. Teaching you how to go from zero to hero in 15 minutes.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770479415883001tO49","RegistrantKey":"1768245294609build26-1770479415883001tO49"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"57739c3f-a75e-4a04-9364-838db00b4da7","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nikos Delis","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293264752001CkVj","langLocale":"en-US","sessionId":"1776293264752001CkVj","sessionInstanceId":"1776293264752001CkVj","sessionCode":"LTG459","sessionCodeNormalized":"EN-US-LTG459","title":"Anatomy of a cloud based, multi tenant, SaaS product","sortTitle":"anatomy of a cloud based, multi tenant, saas product","description":"In this session, I’ll walk through the real-world process of designing and building a cloud-based, multi-tenant SaaS product on Azure. We’ll explore how we use Azure Service Bus, Azure Functions and Durable Functions, Cosmos DB, Microsoft Foundry, and MCP to build a scalable, maintainable system. Expect a practical, technical breakdown of architecture decisions, real challenges, and lessons you can apply when building modern SaaS solutions on Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:00 - 00:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293264752001CkVj_20260512030219583.avif","speakerIds":["1768245294609build26-1772489713010001FU2l"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.889Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293264752001CkVj_20260512030219583.avif","ogTitle":"Anatomy of a cloud based, multi tenant, SaaS product","ogDescription":"In this session, I’ll walk through the real-world process of designing and building a cloud-based, multi-tenant SaaS product on Azure. We’ll explore how we use Azure Service Bus, Azure Functions and Durable Functions, Cosmos DB, Microsoft Foundry, and MCP to build a scalable, maintainable system. Expect a practical, technical breakdown of architecture decisions, real challenges, and lessons you can apply when building modern SaaS solutions on Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772489713010001FU2l","RegistrantKey":"1768245294609build26-1772489713010001FU2l"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"9fa94bb8-274b-4673-9cd2-5f43ab8307fe","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Cole Morrison","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392203158001IFbh","langLocale":"en-US","sessionId":"1774392203158001IFbh","sessionInstanceId":"1774392203158001IFbh","sessionCode":"LTGSP480","sessionCodeNormalized":"EN-US-LTGSP480","title":"Architect and build cloud infrastructure in the AI Era","sortTitle":"architect and build cloud infrastructure in the ai era","description":"What's changed in cloud infrastructure architecture and builds in the era of AI? Given that AI is rapidly changing everyday life in both consumer and business sectors, there are massive shifts occurring in cloud-based systems as well. In this talk, we'll explore how the patterns, principles, and pieces have changed—or stayed the same—for hosting scalable, production-ready systems in 2026.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392203158001IFbh_20260512023615629.avif","speakerIds":["1768245294609build26-1775526930403001fDTI"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.637Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["HashiCorp","an IBM company"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392203158001IFbh_20260512023615629.avif","ogTitle":"Architect and build cloud infrastructure in the AI Era","ogDescription":"What's changed in cloud infrastructure architecture and builds in the era of AI? Given that AI is rapidly changing everyday life in both consumer and business sectors, there are massive shifts occurring in cloud-based systems as well. In this talk, we'll explore how the patterns, principles, and pieces have changed—or stayed the same—for hosting scalable, production-ready systems in 2026.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775526930403001fDTI","RegistrantKey":"1768245294609build26-1775526930403001fDTI"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"AKS","logicalValue":"AKS"}]},{"@search.score":1.0,"documentId":"88386ff1-d004-448d-b191-604bc7993e3d","startDateTime":"2026-06-02T19:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jamshed Damkewala, michael Gamble, Clay Nelson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392118386001zbpm","langLocale":"en-US","sessionId":"1774392118386001zbpm","sessionInstanceId":"1774392118386001zbpm","sessionCode":"LTGSP479","sessionCodeNormalized":"EN-US-LTGSP479","title":"Automate Windows on Arm migrations with AI-assisted specs","sortTitle":"automate windows on arm migrations with ai-assisted specs","description":"Porting Windows apps to Arm used to take weeks: identifying dependencies, chasing compatibility gaps, and relying on trial and error. With AI, the same workflow now fits in a single session. This session shows Windows developers a spec-driven, AI-assisted workflow where a migration spec defines the architecture, GitHub Copilot applies Arm-aware changes, and GitHub Actions with WoA runners validate builds in CI. You’ll leave with a workflow that transforms your next migration from project to PR.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:40 - 19:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392118386001zbpm_20260519172751516.avif","speakerIds":["1768245294609build26-1774284985889001po0v","1768245294609build26-1776815165834001kPZC","1768245294609build26-1776816865122001viae"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.637Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Arm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392118386001zbpm_20260519172751516.avif","ogTitle":"Automate Windows on Arm migrations with AI-assisted specs","ogDescription":"Porting Windows apps to Arm used to take weeks: identifying dependencies, chasing compatibility gaps, and relying on trial and error. With AI, the same workflow now fits in a single session. This session shows Windows developers a spec-driven, AI-assisted workflow where a migration spec defines the architecture, GitHub Copilot applies Arm-aware changes, and GitHub Actions with WoA runners validate builds in CI. You’ll leave with a workflow that transforms your next migration from project to PR.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776816865122001viae","RegistrantKey":"1768245294609build26-1776816865122001viae"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284985889001po0v","RegistrantKey":"1768245294609build26-1774284985889001po0v"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776815165834001kPZC","RegistrantKey":"1768245294609build26-1776815165834001kPZC"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Cost Management","logicalValue":"Cost Management"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"SAP","logicalValue":"SAP"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"MongoDB Migration","logicalValue":"MongoDB Migration"},{"displayValue":"Oracle Migration","logicalValue":"Oracle Migration"},{"displayValue":"On-prem Migration","logicalValue":"On-prem Migration"},{"displayValue":"MySQL Migration","logicalValue":"MySQL Migration"},{"displayValue":"Postgres Migration","logicalValue":"Postgres Migration"},{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"Azure Database for MySQL","logicalValue":"Azure Database for MySQL"},{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"SQL Server Containers on Kubernetes","logicalValue":"SQL Server Containers on Kubernetes"},{"displayValue":"CosmosDB","logicalValue":"CosmosDB"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"GitHub Advanced Security","logicalValue":"GitHub Advanced Security"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Software Development Company","logicalValue":"Software Development Company"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"AKS","logicalValue":"AKS"},{"displayValue":"Azure Copilot","logicalValue":"Azure Copilot"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"Azure Linux","logicalValue":"Azure Linux"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"OSS CI/CD Libraries","logicalValue":"OSS CI/CD Libraries"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"551254c5-da9f-45b3-ab11-877b079f9739","startDateTime":"2026-06-03T18:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nandini Muralidharan, Ashish Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018712070001rAew","langLocale":"en-US","sessionId":"1774018712070001rAew","sessionInstanceId":"1774018712070001rAew","sessionCode":"LTG405","sessionCodeNormalized":"EN-US-LTG405","title":"Better tests, faster. GitHub Copilot does the heavy lifting","sortTitle":"better tests, faster. github copilot does the heavy lifting","description":"AI agents write code faster than ever — but quality still matters. In this demo, see how GitHub Copilot transforms your testing workflow — from generating test suites to TDD-style agent development, catching edge cases, and validating agent-generated code before it merges. From unit tests to load testing, whether retrofitting a legacy codebase or building greenfield, see the patterns that keep AI-assisted development reliable and production-ready.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:10 - 18:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018712070001rAew_20260518110540709.avif","speakerIds":["1768245294609build26-1776202191633001tt4K","1768245294609build26-1778747160284001CBC0"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.851Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018712070001rAew_20260518110540709.avif","ogTitle":"Better tests, faster. GitHub Copilot does the heavy lifting","ogDescription":"AI agents write code faster than ever — but quality still matters. In this demo, see how GitHub Copilot transforms your testing workflow — from generating test suites to TDD-style agent development, catching edge cases, and validating agent-generated code before it merges. From unit tests to load testing, whether retrofitting a legacy codebase or building greenfield, see the patterns that keep AI-assisted development reliable and production-ready.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776202191633001tt4K","RegistrantKey":"1768245294609build26-1776202191633001tt4K"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778747160284001CBC0","RegistrantKey":"1768245294609build26-1778747160284001CBC0"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"82ff64a4-7787-4796-a5b0-752892531340","startDateTime":"2026-06-04T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nandini Muralidharan, Ashish Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778263360832001mLJi","langLocale":"en-US","sessionId":"1778263360832001mLJi","sessionInstanceId":"1778263360832001mLJi","sessionCode":"LTG405-R1","sessionCodeNormalized":"EN-US-LTG405-R1","title":"Better tests, faster. GitHub Copilot does the heavy lifting (Repeat)","sortTitle":"better tests, faster. github copilot does the heavy lifting (repeat)","description":"AI agents write code faster than ever — but quality still matters. In this demo, see how GitHub Copilot transforms your testing workflow — from generating test suites to TDD-style agent development, catching edge cases, and validating agent-generated code before it merges. From unit tests to load testing, whether retrofitting a legacy codebase or building greenfield, see the patterns that keep AI-assisted development reliable and production-ready.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263360832001mLJi_20260518110808115.avif","speakerIds":["1768245294609build26-1776202191633001tt4K","1768245294609build26-1778747160284001CBC0"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.776Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263360832001mLJi_20260518110808115.avif","ogTitle":"Better tests, faster. GitHub Copilot does the heavy lifting (Repeat)","ogDescription":"AI agents write code faster than ever — but quality still matters. In this demo, see how GitHub Copilot transforms your testing workflow — from generating test suites to TDD-style agent development, catching edge cases, and validating agent-generated code before it merges. From unit tests to load testing, whether retrofitting a legacy codebase or building greenfield, see the patterns that keep AI-assisted development reliable and production-ready.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776202191633001tt4K","RegistrantKey":"1768245294609build26-1776202191633001tt4K"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778747160284001CBC0","RegistrantKey":"1768245294609build26-1778747160284001CBC0"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"7fe7a41f-59f1-4aea-99bf-30fb853c0c04","startDateTime":"2026-06-03T17:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Sauber","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776294028217001xLPG","langLocale":"en-US","sessionId":"1776294028217001xLPG","sessionInstanceId":"1776294028217001xLPG","sessionCode":"LTG465","sessionCodeNormalized":"EN-US-LTG465","title":"Beyond the Basics: What GitHub Can Do Today","sortTitle":"beyond the basics: what github can do today","description":"GitHub has been the developer platform of choice for well over a decade. During that time, GitHub continues to add new features and often times people don't realize what's out there. This lightning talk will rapid fire through 10 different things GitHub can do that you might not have seen before. We'll cover AI, Agentic Workflows, Actions, Code Collaboration and more. By the end of this lightning talk, you'll walk away with a number of practical ideas to take back to apply at work immediately.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:10 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776294028217001xLPG_20260512030331182.avif","speakerIds":["1768245294609build26-1776294077207001CQFi"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.89Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776294028217001xLPG_20260512030331182.avif","ogTitle":"Beyond the Basics: What GitHub Can Do Today","ogDescription":"GitHub has been the developer platform of choice for well over a decade. During that time, GitHub continues to add new features and often times people don't realize what's out there. This lightning talk will rapid fire through 10 different things GitHub can do that you might not have seen before. We'll cover AI, Agentic Workflows, Actions, Code Collaboration and more. By the end of this lightning talk, you'll walk away with a number of practical ideas to take back to apply at work immediately.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776294077207001CQFi","RegistrantKey":"1768245294609build26-1776294077207001CQFi"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"69762a65-d974-4105-a3c3-3ea4457597cf","startDateTime":"2026-06-03T16:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Hong Gao, Matt McCleary","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717544001rIrk","langLocale":"en-US","sessionId":"1774018717544001rIrk","sessionInstanceId":"1774018717544001rIrk","sessionCode":"LTG429","sessionCodeNormalized":"EN-US-LTG429","title":"Broken, costly? Debug and operate AI agents with Azure Monitor","sortTitle":"broken, costly? debug and operate ai agents with azure monitor","description":"In this 15 minute session, we'll show how to debug and optimize your production agents. Start by instrumenting your agent with Application Insights instrumentation MCP. Then see it pay off in a new Agent Fleet view, where you can spot regressions across your subscription, agentic transaction maps, ROI and cost dashboards. Now you can move you from \"what happened?\" to \"what do I do next?\" Same OTel trace lights up Azure Monitor and Foundry for evals and human in the loop feedback.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:50 - 17:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717544001rIrk_20260518110613417.avif","speakerIds":["1768245294609build26-1778644007428001P51T","1768245294609build26-1778644324520001gpJ1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.857Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717544001rIrk_20260518110613417.avif","ogTitle":"Broken, costly? Debug and operate AI agents with Azure Monitor","ogDescription":"In this 15 minute session, we'll show how to debug and optimize your production agents. Start by instrumenting your agent with Application Insights instrumentation MCP. Then see it pay off in a new Agent Fleet view, where you can spot regressions across your subscription, agentic transaction maps, ROI and cost dashboards. Now you can move you from \"what happened?\" to \"what do I do next?\" Same OTel trace lights up Azure Monitor and Foundry for evals and human in the loop feedback.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778644007428001P51T","RegistrantKey":"1768245294609build26-1778644007428001P51T"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778644324520001gpJ1","RegistrantKey":"1768245294609build26-1778644324520001gpJ1"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}]},{"@search.score":1.0,"documentId":"80b44f33-cde1-4d26-aa44-1fad999830a9","startDateTime":"2026-06-02T21:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Greg Crist","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775146351584001ha83","langLocale":"en-US","sessionId":"1775146351584001ha83","sessionInstanceId":"1775146351584001ha83","sessionCode":"LTGSP490","sessionCodeNormalized":"EN-US-LTGSP490","title":"Build AI agents with context engineering for autonomous execution","sortTitle":"build ai agents with context engineering for autonomous execution","description":"High impact AI agents require precision search. While models get the glory, context engineering is the real work.This live demo shows how you can ship context driven agents using Elastic Agent Builder on Azure. Master context engineering with Elasticsearch hybrid search via Index Search and ES|QL. Use Elastic Workflows to move from Q&A to autonomous action and leverage built-in MCP to connect to the Microsoft ecosystem. Deploy a production ready reference implementation today.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:50 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1775146351584001ha83_20260512024651908.avif","speakerIds":["1768245294609build26-1764099185727001ifAC"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.845Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Elastic"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1775146351584001ha83_20260512024651908.avif","ogTitle":"Build AI agents with context engineering for autonomous execution","ogDescription":"High impact AI agents require precision search. While models get the glory, context engineering is the real work.This live demo shows how you can ship context driven agents using Elastic Agent Builder on Azure. Master context engineering with Elasticsearch hybrid search via Index Search and ES|QL. Use Elastic Workflows to move from Q&A to autonomous action and leverage built-in MCP to connect to the Microsoft ecosystem. Deploy a production ready reference implementation today.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1764099185727001ifAC","RegistrantKey":"1768245294609build26-1764099185727001ifAC"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Entra","logicalValue":"Entra"}]},{"@search.score":1.0,"documentId":"6358b69e-1b66-4964-97c9-c4df7a836369","startDateTime":"2026-06-03T23:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prashant G Bhoyar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779814137678001ohCu","langLocale":"en-US","sessionId":"1779814137678001ohCu","sessionInstanceId":"1779814137678001ohCu","sessionCode":"LTG448-R1","sessionCodeNormalized":"EN-US-LTG448-R1","title":"Build Intelligent Agents with Work IQ, Foundry IQ, and Fabric IQ","sortTitle":"build intelligent agents with work iq, foundry iq, and fabric iq","description":"What if your AI agents truly understood how your business works, had instant access to trusted enterprise data, and could take meaningful action from end to end? In this demo-driven session, we’ll cover how to build enterprise-ready AI agents that understand business context, use trusted data, and take real action. In this fast-paced session, we’ll bring Work IQ, Foundry IQ, and Fabric IQ together with Copilot Studio through best practices and live demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:40 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814137678001ohCu_20260527112740581.avif","speakerIds":["1768245294609build26-1771723064928001lcoH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.239Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814137678001ohCu_20260527112740581.avif","ogTitle":"Build Intelligent Agents with Work IQ, Foundry IQ, and Fabric IQ","ogDescription":"What if your AI agents truly understood how your business works, had instant access to trusted enterprise data, and could take meaningful action from end to end? In this demo-driven session, we’ll cover how to build enterprise-ready AI agents that understand business context, use trusted data, and take real action. In this fast-paced session, we’ll bring Work IQ, Foundry IQ, and Fabric IQ together with Copilot Studio through best practices and live demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771723064928001lcoH","RegistrantKey":"1768245294609build26-1771723064928001lcoH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"b91cf5a9-7259-4dac-adf3-886a9a8baf88","startDateTime":"2026-06-03T19:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Prashant G Bhoyar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776291995849001LVw2","langLocale":"en-US","sessionId":"1776291995849001LVw2","sessionInstanceId":"1776291995849001LVw2","sessionCode":"LTG448","sessionCodeNormalized":"EN-US-LTG448","title":"Build Intelligent Agents with Work IQ, Foundry IQ, and Fabric IQ","sortTitle":"build intelligent agents with work iq, foundry iq, and fabric iq","description":"What if your AI agents truly understood how your business works, had instant access to trusted enterprise data, and could take meaningful action from end to end? In this demo-driven session, we’ll cover how to build enterprise-ready AI agents that understand business context, use trusted data, and take real action. In this fast-paced session, we’ll bring Work IQ, Foundry IQ, and Fabric IQ together with Copilot Studio through best practices and live demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:50 - 20:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776291995849001LVw2_20260512030002798.avif","speakerIds":["1768245294609build26-1771723064928001lcoH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.888Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776291995849001LVw2_20260512030002798.avif","ogTitle":"Build Intelligent Agents with Work IQ, Foundry IQ, and Fabric IQ","ogDescription":"What if your AI agents truly understood how your business works, had instant access to trusted enterprise data, and could take meaningful action from end to end? In this demo-driven session, we’ll cover how to build enterprise-ready AI agents that understand business context, use trusted data, and take real action. In this fast-paced session, we’ll bring Work IQ, Foundry IQ, and Fabric IQ together with Copilot Studio through best practices and live demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771723064928001lcoH","RegistrantKey":"1768245294609build26-1771723064928001lcoH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"022a49ad-0d78-4a8c-996f-2e5b707247c4","startDateTime":"2026-06-02T21:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Shriram Sankaran","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776266349523001JP6P","langLocale":"en-US","sessionId":"1776266349523001JP6P","sessionInstanceId":"1776266349523001JP6P","sessionCode":"LTGSP481","sessionCodeNormalized":"EN-US-LTGSP481","title":"Build reliable coding agents with .NET and durable workflows","sortTitle":"build reliable coding agents with .net and durable workflows","description":"Modern AI can generate code—but building agents that autonomously fix bugs, run tests, and create pull requests requires real systems engineering. Dive into a hands-on architecture for coding agents built with .NET Web APIs, Temporal workflows, and AI-powered workers. See how to orchestrate long‑running tasks, manage failures, and securely execute code in sandboxed environments to deliver production‑ready autonomous agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:10 - 21:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776266349523001JP6P_20260512025640005.avif","speakerIds":["1768245294609build26-1774530974341001La1w"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.903Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Syncfusion"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776266349523001JP6P_20260512025640005.avif","ogTitle":"Build reliable coding agents with .NET and durable workflows","ogDescription":"Modern AI can generate code—but building agents that autonomously fix bugs, run tests, and create pull requests requires real systems engineering. Dive into a hands-on architecture for coding agents built with .NET Web APIs, Temporal workflows, and AI-powered workers. See how to orchestrate long‑running tasks, manage failures, and securely execute code in sandboxed environments to deliver production‑ready autonomous agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774530974341001La1w","RegistrantKey":"1768245294609build26-1774530974341001La1w"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"0d0f3620-64fd-411f-b105-d794ef436683","startDateTime":"2026-06-03T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"ILIYAS Chawdhary, Vinay Nalam","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774050805352001dHyf","langLocale":"en-US","sessionId":"1774050805352001dHyf","sessionInstanceId":"1774050805352001dHyf","sessionCode":"LTG439","sessionCodeNormalized":"EN-US-LTG439","title":"Building a voice-first agent","sortTitle":"building a voice-first agent","description":"Build your own version of a web-based, voice-first agent using reusable templates that encapsulate proven patterns for conversational UX, prompt design, and content grounding. In this session, you’ll learn how to build an agentic coaching assistant that combines persona, prompt engineering, and curated knowledge. Those same building blocks are now available as templates that you can customize with your own data and use cases. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:40 - 21:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050805352001dHyf_20260512023236390.avif","speakerIds":["1768245294609build26-1775501536483001qQn2","1768245294609build26-1776708589322001Kj9f"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.633Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050805352001dHyf_20260512023236390.avif","ogTitle":"Building a voice-first agent","ogDescription":"Build your own version of a web-based, voice-first agent using reusable templates that encapsulate proven patterns for conversational UX, prompt design, and content grounding. In this session, you’ll learn how to build an agentic coaching assistant that combines persona, prompt engineering, and curated knowledge. Those same building blocks are now available as templates that you can customize with your own data and use cases. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776708589322001Kj9f","RegistrantKey":"1768245294609build26-1776708589322001Kj9f"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775501536483001qQn2","RegistrantKey":"1768245294609build26-1775501536483001qQn2"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"ab0cb204-eba8-4993-8a6a-d8e47e37db1f","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maki Nagase","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774051400990001kVCn","langLocale":"en-US","sessionId":"1774051400990001kVCn","sessionInstanceId":"1774051400990001kVCn","sessionCode":"LTG469","sessionCodeNormalized":"EN-US-LTG469","title":"Building an AI Robot 'StackChan' with GitHub Copilot and Foundry","sortTitle":"building an ai robot 'stackchan' with github copilot and foundry","description":"Physical AI is one of the most exciting frontiers in technology today, but for many developers, building AI-powered robots still feels out of reach. In this 15-minute session, I'll share my hands-on journey as an embedded development beginner who built StackChan — a palm-sized, open-source AI robot — using GitHub Copilot, Microsoft Foundry (gpt-realtime), and .NET.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:30 - 17:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051400990001kVCn_20260512023351978.avif","speakerIds":["1768245294609build26-1773319382603001KpW5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.634Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051400990001kVCn_20260512023351978.avif","ogTitle":"Building an AI Robot 'StackChan' with GitHub Copilot and Foundry","ogDescription":"Physical AI is one of the most exciting frontiers in technology today, but for many developers, building AI-powered robots still feels out of reach. In this 15-minute session, I'll share my hands-on journey as an embedded development beginner who built StackChan — a palm-sized, open-source AI robot — using GitHub Copilot, Microsoft Foundry (gpt-realtime), and .NET.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773319382603001KpW5","RegistrantKey":"1768245294609build26-1773319382603001KpW5"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"5ff10d45-fb83-4d35-8113-230c342fc1f9","startDateTime":"2026-06-03T23:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ehsan Eskandari","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779814253531001gx5R","langLocale":"en-US","sessionId":"1779814253531001gx5R","sessionInstanceId":"1779814253531001gx5R","sessionCode":"LTG453-R1","sessionCodeNormalized":"EN-US-LTG453-R1","title":"Building an End‑to‑End Enterprise AI Platform on Azure","sortTitle":"building an end‑to‑end enterprise ai platform on azure","description":"Enterprise AI systems span ingestion, embeddings, vector search, APIs, and partner integrations—often across serverless and container platforms. In this demo, see how Contoso builds a production‑ready, end‑to‑end AI solution on Azure using Azure Functions, Container Apps, AKS, Cosmos DB vector search. Learn proven patterns for scalability, observability, and secure operations with minimal dev effort.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:40 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814253531001gx5R_20260527112806531.avif","speakerIds":["1768245294609build26-1757087809738001nkEU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.239Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814253531001gx5R_20260527112806531.avif","ogTitle":"Building an End‑to‑End Enterprise AI Platform on Azure","ogDescription":"Enterprise AI systems span ingestion, embeddings, vector search, APIs, and partner integrations—often across serverless and container platforms. In this demo, see how Contoso builds a production‑ready, end‑to‑end AI solution on Azure using Azure Functions, Container Apps, AKS, Cosmos DB vector search. Learn proven patterns for scalability, observability, and secure operations with minimal dev effort.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1757087809738001nkEU","RegistrantKey":"1768245294609build26-1757087809738001nkEU"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"c72cc653-1be1-47d6-a2a2-b0c0dfad8ce6","startDateTime":"2026-06-02T20:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ehsan Eskandari","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776292817165001Xnbo","langLocale":"en-US","sessionId":"1776292817165001Xnbo","sessionInstanceId":"1776292817165001Xnbo","sessionCode":"LTG453","sessionCodeNormalized":"EN-US-LTG453","title":"Building an End‑to‑End Enterprise AI Platform on Azure","sortTitle":"building an end‑to‑end enterprise ai platform on azure","description":"Enterprise AI systems span ingestion, embeddings, vector search, APIs, and partner integrations—often across serverless and container platforms. In this demo, see how Contoso builds a production‑ready, end‑to‑end AI solution on Azure using Azure Functions, Container Apps, AKS, Cosmos DB vector search. Learn proven patterns for scalability, observability, and secure operations with minimal dev effort.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:50 - 21:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292817165001Xnbo_20260512030056762.avif","speakerIds":["1768245294609build26-1757087809738001nkEU"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.888Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292817165001Xnbo_20260512030056762.avif","ogTitle":"Building an End‑to‑End Enterprise AI Platform on Azure","ogDescription":"Enterprise AI systems span ingestion, embeddings, vector search, APIs, and partner integrations—often across serverless and container platforms. In this demo, see how Contoso builds a production‑ready, end‑to‑end AI solution on Azure using Azure Functions, Container Apps, AKS, Cosmos DB vector search. Learn proven patterns for scalability, observability, and secure operations with minimal dev effort.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1757087809738001nkEU","RegistrantKey":"1768245294609build26-1757087809738001nkEU"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"58301532-e591-48e5-abd6-18adeb8516f8","startDateTime":"2026-06-03T21:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dakota Shafer, Taiki Yoshida, Laura Wolf","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703178001rqXS","langLocale":"en-US","sessionId":"1774018703178001rqXS","sessionInstanceId":"1774018703178001rqXS","sessionCode":"LTG423","sessionCodeNormalized":"EN-US-LTG423","title":"Building end-to-end agentic automation in 15 mins","sortTitle":"building end-to-end agentic automation in 15 mins","description":"In just 15 minutes, see how Copilot Studio can deliver a complete autonomous agentic automation —from idea to deployment. Learn to use a combination of out of the box and your own custom MCPs, leverage Copilot Studio’s both goal-based automations like orchestration and CUA as well as rule-based automation capabilities like Agent Flows. This session reveals how to streamline tasks, reduce manual effort, and supercharge productivity for developers wanting to modernize their internal systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:20 - 21:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703178001rqXS_20260512153016845.avif","speakerIds":["1768245294609build26-1777666248425001Pf6C","1768245294609build26-1751039983976001aUFZ","1768245294609build26-1767984116727001aLg7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.458Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703178001rqXS_20260512153016845.avif","ogTitle":"Building end-to-end agentic automation in 15 mins","ogDescription":"In just 15 minutes, see how Copilot Studio can deliver a complete autonomous agentic automation —from idea to deployment. Learn to use a combination of out of the box and your own custom MCPs, leverage Copilot Studio’s both goal-based automations like orchestration and CUA as well as rule-based automation capabilities like Agent Flows. This session reveals how to streamline tasks, reduce manual effort, and supercharge productivity for developers wanting to modernize their internal systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Content Lead","SpeakerId":"1768245294609build26-1767984116727001aLg7","RegistrantKey":"1768245294609build26-1767984116727001aLg7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777666248425001Pf6C","RegistrantKey":"1768245294609build26-1777666248425001Pf6C"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751039983976001aUFZ","RegistrantKey":"1768245294609build26-1751039983976001aUFZ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Copilot Studio","logicalValue":"Copilot Studio"}]},{"@search.score":1.0,"documentId":"4b2a25b1-812e-4bf1-8106-e255ae12bd71","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Stuart van der Lee","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776294107511001Z3tj","langLocale":"en-US","sessionId":"1776294107511001Z3tj","sessionInstanceId":"1776294107511001Z3tj","sessionCode":"LTG466","sessionCodeNormalized":"EN-US-LTG466","title":"Building your own MCP server","sortTitle":"building your own mcp server","description":"The Model Context Protocol (MCP) is all the rage and rightly so. As the USB-C for AI, it is an incredibly powerful tool that acts as a bridge between real-world systems and LLMs. But how do you get started with MCP and more importantly: how do you build your own MCP server? In this session we'll build an MCP server from scratch using the MCP C# SDK. We'll look at how to build tools, prompts and resources and how to leverage these with our own AI agent in Microsoft Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:30 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776294107511001Z3tj_20260512030339887.avif","speakerIds":["1768245294609build26-1770898149489001eL44"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.89Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776294107511001Z3tj_20260512030339887.avif","ogTitle":"Building your own MCP server","ogDescription":"The Model Context Protocol (MCP) is all the rage and rightly so. As the USB-C for AI, it is an incredibly powerful tool that acts as a bridge between real-world systems and LLMs. But how do you get started with MCP and more importantly: how do you build your own MCP server? In this session we'll build an MCP server from scratch using the MCP C# SDK. We'll look at how to build tools, prompts and resources and how to leverage these with our own AI agent in Microsoft Foundry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770898149489001eL44","RegistrantKey":"1768245294609build26-1770898149489001eL44"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"78a74db2-85d3-4483-90ac-bc3f67e85298","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Brooks Naylor","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888591641001IgLK","langLocale":"en-US","sessionId":"1776888591641001IgLK","sessionInstanceId":"1776888591641001IgLK","sessionCode":"LTGSP496","sessionCodeNormalized":"EN-US-LTGSP496","title":"Can LLMs generate high quality code?","sortTitle":"can llms generate high quality code?","description":"AI tools are quickly becoming key contributors to the output of software engineering teams, but the code they write needs to be reliable, maintainable, and secure. Not all LLMs deliver that. Issues can also compound, and cause downstream problems. Our deep evaluation of LLMs surfaces their strengths and weaknesses. Learn how to pair AI agents with deterministic analysis to allow you to manage quality and security risks without giving up the productivity gains that made AI worth adopting.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:00 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888591641001IgLK_20260528122912599.avif","speakerIds":["1768245294609build26-1778602617292001BArg"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.386Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Sonar"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888591641001IgLK_20260528122912599.avif","ogTitle":"Can LLMs generate high quality code?","ogDescription":"AI tools are quickly becoming key contributors to the output of software engineering teams, but the code they write needs to be reliable, maintainable, and secure. Not all LLMs deliver that. Issues can also compound, and cause downstream problems. Our deep evaluation of LLMs surfaces their strengths and weaknesses. Learn how to pair AI agents with deterministic analysis to allow you to manage quality and security risks without giving up the productivity gains that made AI worth adopting.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778602617292001BArg","RegistrantKey":"1768245294609build26-1778602617292001BArg"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"29043fce-5e8a-45cb-ad6b-5411170106a5","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Carie Fisher, Aaron Gustafson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774051494766001FDla","langLocale":"en-US","sessionId":"1774051494766001FDla","sessionInstanceId":"1774051494766001FDla","sessionCode":"LTG470","sessionCodeNormalized":"EN-US-LTG470","title":"Can your AI pass the accessibility test?","sortTitle":"can your ai pass the accessibility test?","description":"As AI speeds up development, accessibility must be built in from the start. This lightning talk highlights GitHub’s Accessibility Scanner and how it brings accessibility signals directly into developer workflows, alongside security and quality checks. We’ll also touch on the broader state of AI and Accessibility at Microsoft, focusing on how accessibility is being embedded into AI tooling and engineering practices so innovation scales inclusion.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:30 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051494766001FDla_20260512023404154.avif","speakerIds":["1768245294609build26-1777300452434001pMw7","1768245294609build26-1773332637431001OOg9"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.634Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051494766001FDla_20260512023404154.avif","ogTitle":"Can your AI pass the accessibility test?","ogDescription":"As AI speeds up development, accessibility must be built in from the start. This lightning talk highlights GitHub’s Accessibility Scanner and how it brings accessibility signals directly into developer workflows, alongside security and quality checks. We’ll also touch on the broader state of AI and Accessibility at Microsoft, focusing on how accessibility is being embedded into AI tooling and engineering practices so innovation scales inclusion.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773332637431001OOg9","RegistrantKey":"1768245294609build26-1773332637431001OOg9"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777300452434001pMw7","RegistrantKey":"1768245294609build26-1777300452434001pMw7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"e2807b06-9bcd-4fc9-ac15-4fbcc0ed2051","startDateTime":"2026-06-02T23:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Aman Singh","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018708081001rhQb","langLocale":"en-US","sessionId":"1774018708081001rhQb","sessionInstanceId":"1774018708081001rhQb","sessionCode":"LTG414","sessionCodeNormalized":"EN-US-LTG414","title":"Change-driven AI agents with Drasi and LangChain","sortTitle":"change-driven ai agents with drasi and langchain","description":"Drasi is a change‑detection platform for databases that enables reactive, change‑driven AI agents. The LangChain‑Drasi integration connects real‑time data changes to agent workflows using continuous queries. Agents can discover and subscribe to changes, update state and memory, and adapt execution—turning static agents into long‑lived, responsive systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:10 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018708081001rhQb_20260512021913750.avif","speakerIds":["1768245294609build26-1775094106472001eWz6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018708081001rhQb_20260512021913750.avif","ogTitle":"Change-driven AI agents with Drasi and LangChain","ogDescription":"Drasi is a change‑detection platform for databases that enables reactive, change‑driven AI agents. The LangChain‑Drasi integration connects real‑time data changes to agent workflows using continuous queries. Agents can discover and subscribe to changes, update state and memory, and adapt execution—turning static agents into long‑lived, responsive systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775094106472001eWz6","RegistrantKey":"1768245294609build26-1775094106472001eWz6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"OSS","logicalValue":"OSS"}]},{"@search.score":1.0,"documentId":"1e3d9d16-0442-450b-ad6d-ead2fa29c72e","startDateTime":"2026-06-03T16:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Priyanka Shah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776292232227001Ltui","langLocale":"en-US","sessionId":"1776292232227001Ltui","sessionInstanceId":"1776292232227001Ltui","sessionCode":"LTG449","sessionCodeNormalized":"EN-US-LTG449","title":"Closing the Loop: Feedback based Learning for Agentic RAG on Azure","sortTitle":"closing the loop: feedback based learning for agentic rag on azure","description":"Search and retrieval quality is the foundation of effective agentic RAG. This session explores practical ways to improve RAG accuracy using human feedback across Microsoft Foundry and MAF patterns—from relevance rankings, preference signals and error tagging to LLM-as-a-judge, Simulation SDKs, semantic reranking, query refinement, and grounded evaluation. Learn how to turn user interactions into measurable signals that continuously improve retrieval, ranking, and answer quality in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:50 - 17:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292232227001Ltui_20260512030015514.avif","speakerIds":["1768245294609build26-1773185276975001cATe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.889Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292232227001Ltui_20260512030015514.avif","ogTitle":"Closing the Loop: Feedback based Learning for Agentic RAG on Azure","ogDescription":"Search and retrieval quality is the foundation of effective agentic RAG. This session explores practical ways to improve RAG accuracy using human feedback across Microsoft Foundry and MAF patterns—from relevance rankings, preference signals and error tagging to LLM-as-a-judge, Simulation SDKs, semantic reranking, query refinement, and grounded evaluation. Learn how to turn user interactions into measurable signals that continuously improve retrieval, ranking, and answer quality in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773185276975001cATe","RegistrantKey":"1768245294609build26-1773185276975001cATe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"3ba5ee25-2cce-4d6c-ae36-06ff4df233df","startDateTime":"2026-06-02T23:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Denzil Ribeiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707907001rSTJ","langLocale":"en-US","sessionId":"1774018707907001rSTJ","sessionInstanceId":"1774018707907001rSTJ","sessionCode":"LTG413","sessionCodeNormalized":"EN-US-LTG413","title":"Cloud‑Native PostgreSQL, Rebuilt for Scale: Azure HorizonDB","sortTitle":"cloud‑native postgresql, rebuilt for scale: azure horizondb","description":"Modern applications demand elastic scale without complexity. In this session, we explore how Azure HorizonDBre-architects PostgreSQL for cloud-native scalability by decoupling compute, storage, and caching to deliver predictable performance at massive scale. See how it delivers sub millisecond latency, scales out reads and delivers 3x the transactional throughput of self-managed Postgres.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:50 - 00:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707907001rSTJ_20260513202916111.avif","speakerIds":["1768245294609build26-1765804946228001vzRn"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707907001rSTJ_20260513202916111.avif","ogTitle":"Cloud‑Native PostgreSQL, Rebuilt for Scale: Azure HorizonDB","ogDescription":"Modern applications demand elastic scale without complexity. In this session, we explore how Azure HorizonDBre-architects PostgreSQL for cloud-native scalability by decoupling compute, storage, and caching to deliver predictable performance at massive scale. See how it delivers sub millisecond latency, scales out reads and delivers 3x the transactional throughput of self-managed Postgres.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765804946228001vzRn","RegistrantKey":"1768245294609build26-1765804946228001vzRn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"67e8ec95-48cf-419f-85d4-40bf82e9babb","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Patrick Brown","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774050911164001k5R0","langLocale":"en-US","sessionId":"1774050911164001k5R0","sessionInstanceId":"1774050911164001k5R0","sessionCode":"LTG441","sessionCodeNormalized":"EN-US-LTG441","title":"CopyFlag: Imagine Cup Finalists Tackling Creative Protection","sortTitle":"copyflag: imagine cup finalists tackling creative protection","description":"CopyFlag helps independent creators fight AI-enabled infringement by continuously scanning marketplaces and social platforms for remixed copies. Built on Azure AI Vision, Azure OpenAI, Azure AI Search, and Azure Machine Learning, it vectorizes images, scores matches, and automates takedowns in minutes. Multi-model detection, combined with Azure AI Content Safety and offender network mapping, makes creator protection scalable and affordable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:30 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050911164001k5R0_20260512023258507.avif","speakerIds":["1768245294609build26-1776076731812001gK5F"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.633Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050911164001k5R0_20260512023258507.avif","ogTitle":"CopyFlag: Imagine Cup Finalists Tackling Creative Protection","ogDescription":"CopyFlag helps independent creators fight AI-enabled infringement by continuously scanning marketplaces and social platforms for remixed copies. Built on Azure AI Vision, Azure OpenAI, Azure AI Search, and Azure Machine Learning, it vectorizes images, scores matches, and automates takedowns in minutes. Multi-model detection, combined with Azure AI Content Safety and offender network mapping, makes creator protection scalable and affordable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"Node.js","logicalValue":"Node.js"},{"displayValue":"TypeScript","logicalValue":"TypeScript"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776076731812001gK5F","RegistrantKey":"1768245294609build26-1776076731812001gK5F"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Azure ML","logicalValue":"Azure ML"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"PyTorch","logicalValue":"PyTorch"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"User Privacy","logicalValue":"User Privacy"},{"displayValue":"App Integration","logicalValue":"App Integration"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"}]},{"@search.score":1.0,"documentId":"c2206f1c-2936-4c0c-a4f3-1d90a4f2bd3c","startDateTime":"2026-06-03T18:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Hannah Cooper","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018719827001r6Oh","langLocale":"en-US","sessionId":"1774018719827001r6Oh","sessionInstanceId":"1774018719827001r6Oh","sessionCode":"LTG433","sessionCodeNormalized":"EN-US-LTG433","title":"Debugging and Securing agent developer experiences on Windows","sortTitle":"debugging and securing agent developer experiences on windows","description":"This session demonstrates agent developer experiences on Windows. Using live code, we expose Windows capabilities and trace how it is packaged, trusted, and executed by the platform. You will learn how Windows enforces security and lifecycle boundaries, what responsibilities stay with the developer, and how to debug common failure modes such as denied access and misconfiguration.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:30 - 18:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018719827001r6Oh_20260512153054115.avif","speakerIds":["1768245294609build26-1758133727495001O3tv"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.859Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018719827001r6Oh_20260512153054115.avif","ogTitle":"Debugging and Securing agent developer experiences on Windows","ogDescription":"This session demonstrates agent developer experiences on Windows. Using live code, we expose Windows capabilities and trace how it is packaged, trusted, and executed by the platform. You will learn how Windows enforces security and lifecycle boundaries, what responsibilities stay with the developer, and how to debug common failure modes such as denied access and misconfiguration.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758133727495001O3tv","RegistrantKey":"1768245294609build26-1758133727495001O3tv"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"On-Device Registry","logicalValue":"On-Device Registry"}]},{"@search.score":1.0,"documentId":"978daa44-028e-47d7-b269-bf8854aa9d42","startDateTime":"2026-06-03T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Steve Pousty","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888818056001rfSc","langLocale":"en-US","sessionId":"1776888818056001rfSc","sessionInstanceId":"1776888818056001rfSc","sessionCode":"LTGSP494","sessionCodeNormalized":"EN-US-LTGSP494","title":"Deploy open AI workloads on Azure with Red Hat and AKS","sortTitle":"deploy open ai workloads on azure with red hat and aks","description":"Discover how Red Hat brings open choice to Azure AI development—from OS to inference. Explore how Red Hat AI Inference delivers a consistent, portable open-source stack on AKS. See how MCP and Ansible Lightspeed with Copilot connect AI agents to Azure infrastructure. Whether optimizing LLM throughput or building agents, learn to create flexible AI solutions with Red Hat and Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:40 - 21:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888818056001rfSc_20260602111615995.avif","speakerIds":["1768245294609build26-1780326807888001GNVh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.387Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Red Hat"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888818056001rfSc_20260602111615995.avif","ogTitle":"Deploy open AI workloads on Azure with Red Hat and AKS","ogDescription":"Discover how Red Hat brings open choice to Azure AI development—from OS to inference. Explore how Red Hat AI Inference delivers a consistent, portable open-source stack on AKS. See how MCP and Ansible Lightspeed with Copilot connect AI agents to Azure infrastructure. Whether optimizing LLM throughput or building agents, learn to create flexible AI solutions with Red Hat and Azure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1780326807888001GNVh","RegistrantKey":"1768245294609build26-1780326807888001GNVh"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"1a6f460e-4bfb-4d9a-9533-3a7bc20085c2","startDateTime":"2026-06-02T23:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Freek Berson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776292893312001WFRO","langLocale":"en-US","sessionId":"1776292893312001WFRO","sessionInstanceId":"1776292893312001WFRO","sessionCode":"LTG454","sessionCodeNormalized":"EN-US-LTG454","title":"Deploying Infrastructure and Turning on the Light: Bicep Beyond Azure","sortTitle":"deploying infrastructure and turning on the light: bicep beyond azure","description":"Take your Infrastructure as Code workflows to the next level with Bicep Local Deploy. In this demo-driven session, you’ll learn how to extend Bicep beyond Azure resources and treat external systems as declarative infrastructure. We walk through a custom Bicep extension in .NET that exposes new resource types and integrates with external REST APIs. A memorable live demo with practical insights! Leave with both the knowledge and the code needed to start building your own Bicep extensions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:50 - 00:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292893312001WFRO_20260512030109765.avif","speakerIds":["1768245294609build26-1776289702978001L6Ff"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.887Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292893312001WFRO_20260512030109765.avif","ogTitle":"Deploying Infrastructure and Turning on the Light: Bicep Beyond Azure","ogDescription":"Take your Infrastructure as Code workflows to the next level with Bicep Local Deploy. In this demo-driven session, you’ll learn how to extend Bicep beyond Azure resources and treat external systems as declarative infrastructure. We walk through a custom Bicep extension in .NET that exposes new resource types and integrates with external REST APIs. A memorable live demo with practical insights! Leave with both the knowledge and the code needed to start building your own Bicep extensions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776289702978001L6Ff","RegistrantKey":"1768245294609build26-1776289702978001L6Ff"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Win32 Developers","logicalValue":"Win32 Developers"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"6073a29b-a0ff-40b6-8d4b-c36ad89f3074","startDateTime":"2026-06-03T23:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rufus Harvey","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888738206001CKc6","langLocale":"en-US","sessionId":"1776888738206001CKc6","sessionInstanceId":"1776888738206001CKc6","sessionCode":"LTGSP499","sessionCodeNormalized":"EN-US-LTGSP499","title":"Design efficient AI agents with modular skills, not context bloat","sortTitle":"design efficient ai agents with modular skills, not context bloat","description":"Your agents are getting smarter — and slower, more expensive, and less reliable. Why? Because we've been bolting capabilities onto context windows instead of giving agents what engineers have always given systems: modular, on-demand expertise. In this session with CDW, learn how skills cut agent token usage, drop latency, and lift task accuracy in production — with concrete numbers, reusable patterns, and a checklist you can apply immediately.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:40 - 23:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888738206001CKc6_20260521160841641.avif","speakerIds":["1768245294609build26-1779120859418001UK52"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.386Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["CDW"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888738206001CKc6_20260521160841641.avif","ogTitle":"Design efficient AI agents with modular skills, not context bloat","ogDescription":"Your agents are getting smarter — and slower, more expensive, and less reliable. Why? Because we've been bolting capabilities onto context windows instead of giving agents what engineers have always given systems: modular, on-demand expertise. In this session with CDW, learn how skills cut agent token usage, drop latency, and lift task accuracy in production — with concrete numbers, reusable patterns, and a checklist you can apply immediately.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779120859418001UK52","RegistrantKey":"1768245294609build26-1779120859418001UK52"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"685529ab-59ff-4306-990b-c2f20481cba1","startDateTime":"2026-06-03T21:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Shijo Joy, Edo Segal, Gillian Sheldon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775841352678001DgCO","langLocale":"en-US","sessionId":"1775841352678001DgCO","sessionInstanceId":"1775841352678001DgCO","sessionCode":"LTGSP491","sessionCodeNormalized":"EN-US-LTGSP491","title":"Design multimodal agents that work as collaborators","sortTitle":"design multimodal agents that work as collaborators","description":"What if every person on your team had a dozen competent coworkers available around the clock, across every function, ready to support them regardless of the task? The shift from copilots to coworkers is already underway. Napster Chief Technology and Product Officer Edo Segal will offer a look at what happens when AI actually shows up as part of the team. These multimodal coworkers combine video presence, real-time tool access, and relationship continuity to uplevel the employee experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:00 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1775841352678001DgCO_20260527112953948.avif","speakerIds":["1768245294609build26-1761719658031001RARW","1768245294609build26-1776448795539001JeI6","1768245294609build26-1776448778712001lItD"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.901Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Napster"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1775841352678001DgCO_20260527112953948.avif","ogTitle":"Design multimodal agents that work as collaborators","ogDescription":"What if every person on your team had a dozen competent coworkers available around the clock, across every function, ready to support them regardless of the task? The shift from copilots to coworkers is already underway. Napster Chief Technology and Product Officer Edo Segal will offer a look at what happens when AI actually shows up as part of the team. These multimodal coworkers combine video presence, real-time tool access, and relationship continuity to uplevel the employee experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Owner","SpeakerId":"1768245294609build26-1776448778712001lItD","RegistrantKey":"1768245294609build26-1776448778712001lItD"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776448795539001JeI6","RegistrantKey":"1768245294609build26-1776448795539001JeI6"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1761719658031001RARW","RegistrantKey":"1768245294609build26-1761719658031001RARW"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Windows APIs","logicalValue":"Windows APIs"}]},{"@search.score":1.0,"documentId":"46997ac6-890f-4a3b-bb99-9e3069aac839","startDateTime":"2026-06-02T23:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kendal Van Dyke","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707249001r93j","langLocale":"en-US","sessionId":"1774018707249001r93j","sessionInstanceId":"1774018707249001r93j","sessionCode":"LTG409","sessionCodeNormalized":"EN-US-LTG409","title":"Develop and scale without limits using Azure SQL Database Hyperscale","sortTitle":"develop and scale without limits using azure sql database hyperscale","description":"In this demo‑driven session, see how Azure SQL Database Hyperscale removes scaling limits without changing how you build. Using familiar SQL tools and patterns, we’ll start with a transactional app and show how Hyperscale delivers performance on demand, scales effortlessly as your app grows, and stays AI‑ready by design—so you can focus on writing code, shipping features faster, and building intelligent experiences without managing infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:10 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707249001r93j_20260512021821952.avif","speakerIds":["1768245294609build26-1776186393665001Qsqj"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707249001r93j_20260512021821952.avif","ogTitle":"Develop and scale without limits using Azure SQL Database Hyperscale","ogDescription":"In this demo‑driven session, see how Azure SQL Database Hyperscale removes scaling limits without changing how you build. Using familiar SQL tools and patterns, we’ll start with a transactional app and show how Hyperscale delivers performance on demand, scales effortlessly as your app grows, and stays AI‑ready by design—so you can focus on writing code, shipping features faster, and building intelligent experiences without managing infrastructure.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776186393665001Qsqj","RegistrantKey":"1768245294609build26-1776186393665001Qsqj"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"ed1f663f-d44f-448c-b2a5-70faab73ce86","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Pankaj Bhojwani","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018720119001rG8N","langLocale":"en-US","sessionId":"1774018720119001rG8N","sessionInstanceId":"1774018720119001rG8N","sessionCode":"LTG434","sessionCodeNormalized":"EN-US-LTG434","title":"Elevate your developer productivity with Windows Terminal","sortTitle":"elevate your developer productivity with windows terminal","description":"Developers lose time switching tools to maintain UNIX habits. This session shows how new improvements coming to Windows Terminal support consistent command‑driven workflows so developers can code with reduced context switching. Attendees learn how to create reusable workflows to help build and ship faster using familiar UNIX like workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720119001rG8N_20260512023059271.avif","speakerIds":["1768245294609build26-1776719248150001CLIz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.858Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720119001rG8N_20260512023059271.avif","ogTitle":"Elevate your developer productivity with Windows Terminal","ogDescription":"Developers lose time switching tools to maintain UNIX habits. This session shows how new improvements coming to Windows Terminal support consistent command‑driven workflows so developers can code with reduced context switching. Attendees learn how to create reusable workflows to help build and ship faster using familiar UNIX like workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776719248150001CLIz","RegistrantKey":"1768245294609build26-1776719248150001CLIz"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Windows Development","logicalValue":"Windows Development"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Developer Languages","logicalValue":"Developer Languages"},{"displayValue":"WSL","logicalValue":"WSL"}]},{"@search.score":1.0,"documentId":"4944f79f-0dbf-4cc7-be62-9b1019052bcc","startDateTime":"2026-06-03T22:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Moran Assaf, Meng Tang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703493001r10j","langLocale":"en-US","sessionId":"1774018703493001r10j","sessionInstanceId":"1774018703493001r10j","sessionCode":"LTG425","sessionCodeNormalized":"EN-US-LTG425","title":"Expanding horizons: Foundry Local for devices and on-prem","sortTitle":"expanding horizons: foundry local for devices and on-prem","description":"Discover how Foundry Local extends AI capabilities across multiple platforms in this demo session. Find out how to deploy and optimize models, integrate them with enterprise workflows, and explore best practices for cross-platform compatibility. Learn about real-world use cases that support scalable cloud-to-edge AI solutions from Windows, Mac, and Android, fostering innovation and operational efficiency.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:00 - 22:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703493001r10j_20260515115247078.avif","speakerIds":["1768245294609build26-1778618247524001n8Zk","1768245294609build26-1774383196037001nAji"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.459Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703493001r10j_20260515115247078.avif","ogTitle":"Expanding horizons: Foundry Local for devices and on-prem","ogDescription":"Discover how Foundry Local extends AI capabilities across multiple platforms in this demo session. Find out how to deploy and optimize models, integrate them with enterprise workflows, and explore best practices for cross-platform compatibility. Learn about real-world use cases that support scalable cloud-to-edge AI solutions from Windows, Mac, and Android, fostering innovation and operational efficiency.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774383196037001nAji","RegistrantKey":"1768245294609build26-1774383196037001nAji"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778618247524001n8Zk","RegistrantKey":"1768245294609build26-1778618247524001n8Zk"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Cross-device","logicalValue":"Cross-device"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"1cc9952c-de75-4719-a385-4273bb1bd59e","startDateTime":"2026-06-03T20:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alex Koren, Ryan Whitehead","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888666362001Khir","langLocale":"en-US","sessionId":"1776888666362001Khir","sessionInstanceId":"1776888666362001Khir","sessionCode":"LTGSP495","sessionCodeNormalized":"EN-US-LTGSP495","title":"Extend coding agents across your team with Claude in Microsoft Foundry","sortTitle":"extend coding agents across your team with claude in microsoft foundry","description":"Many teams adopt a coding agent, then ask: why can't the rest of the org work this way? Claude Code handles engineering; Claude Cowork extends the same fundamentals to research, documents, and operational work across organizations. This talk covers how each maps to real enterprise workflows, what's happening inside the task loop, and the setup, auth, and governance patterns that matter when you run them through Microsoft Foundry. Live demos of both, pointed at work your teams already do.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:20 - 20:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888666362001Khir_20260601103610087.avif","speakerIds":["1768245294609build26-1779383763552001b1NK","1768245294609build26-1777992712377001cayS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.385Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Anthropic","PBC"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888666362001Khir_20260601103610087.avif","ogTitle":"Extend coding agents across your team with Claude in Microsoft Foundry","ogDescription":"Many teams adopt a coding agent, then ask: why can't the rest of the org work this way? Claude Code handles engineering; Claude Cowork extends the same fundamentals to research, documents, and operational work across organizations. This talk covers how each maps to real enterprise workflows, what's happening inside the task loop, and the setup, auth, and governance patterns that matter when you run them through Microsoft Foundry. Live demos of both, pointed at work your teams already do.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777992712377001cayS","RegistrantKey":"1768245294609build26-1777992712377001cayS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779383763552001b1NK","RegistrantKey":"1768245294609build26-1779383763552001b1NK"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"fa7dbbdb-83a5-410c-b448-fca2599029e5","startDateTime":"2026-06-02T21:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alastair Fehr, Aarthi Saravanakumar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018717894001rgAt","langLocale":"en-US","sessionId":"1774018717894001rgAt","sessionInstanceId":"1774018717894001rgAt","sessionCode":"LTG431","sessionCodeNormalized":"EN-US-LTG431","title":"Extend your LangChain agent to Agent365: Collaborate like an employee","sortTitle":"extend your langchain agent to agent365: collaborate like an employee","description":"Create a new\u00A0LangChain\u00A0agent, enable Agent 365 observability, and integrate\u00A0Work IQ\u00A0MCP servers. Register the agent with Agent 365 to bring it under a single control plane. We will\u00A0demonstrate\u00A0end-to-end invocation of the\u00A0agent\u00A0by triggering the agent from a comment in a Word document, using\u00A0Work IQ\u00A0MCP servers, and emitting telemetry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:50 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717894001rgAt_20260512022927356.avif","speakerIds":["1768245294609build26-1776362374603001afWn","1768245294609build26-1775086994779001hUjS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.856Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018717894001rgAt_20260512022927356.avif","ogTitle":"Extend your LangChain agent to Agent365: Collaborate like an employee","ogDescription":"Create a new\u00A0LangChain\u00A0agent, enable Agent 365 observability, and integrate\u00A0Work IQ\u00A0MCP servers. Register the agent with Agent 365 to bring it under a single control plane. We will\u00A0demonstrate\u00A0end-to-end invocation of the\u00A0agent\u00A0by triggering the agent from a comment in a Word document, using\u00A0Work IQ\u00A0MCP servers, and emitting telemetry.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775086994779001hUjS","RegistrantKey":"1768245294609build26-1775086994779001hUjS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776362374603001afWn","RegistrantKey":"1768245294609build26-1776362374603001afWn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Work IQ","logicalValue":"Work IQ"}]},{"@search.score":1.0,"documentId":"a7311841-0079-429e-b5d8-805186f014dc","startDateTime":"2026-06-03T23:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Philip Laussermair","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707419001rtDZ","langLocale":"en-US","sessionId":"1774018707419001rtDZ","sessionInstanceId":"1774018707419001rtDZ","sessionCode":"LTG410","sessionCodeNormalized":"EN-US-LTG410","title":"Faster and smarter agents with Redis and Foundry","sortTitle":"faster and smarter agents with redis and foundry","description":"AI agents are only as powerful as the data and memory behind them. In this session, learn how Azure Managed Redis and Microsoft Foundry enable faster, more intelligent agents through high-performance in-memory data, vector search, and semantic caching that deliver the right context in milliseconds while reducing latency and cost. Learn practical patterns—spanning RAG, persistent memory, and event-driven workflows—to build production-ready agents with richer context and real-time reasoning.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:20 - 23:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707419001rtDZ_20260527112846193.avif","speakerIds":["1768245294609build26-1759173805963001hLfB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707419001rtDZ_20260527112846193.avif","ogTitle":"Faster and smarter agents with Redis and Foundry","ogDescription":"AI agents are only as powerful as the data and memory behind them. In this session, learn how Azure Managed Redis and Microsoft Foundry enable faster, more intelligent agents through high-performance in-memory data, vector search, and semantic caching that deliver the right context in milliseconds while reducing latency and cost. Learn practical patterns—spanning RAG, persistent memory, and event-driven workflows—to build production-ready agents with richer context and real-time reasoning.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1759173805963001hLfB","RegistrantKey":"1768245294609build26-1759173805963001hLfB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"68592101-e844-482b-bff5-f4a2291beca8","startDateTime":"2026-06-02T21:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Julia Kasper, Josh Spicer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711903001rZpY","langLocale":"en-US","sessionId":"1774018711903001rZpY","sessionInstanceId":"1774018711903001rZpY","sessionCode":"LTG404","sessionCodeNormalized":"EN-US-LTG404","title":"Fire, forget, sleep. VS Code agents that run without you","sortTitle":"fire, forget, sleep. vs code agents that run without you","description":"You can only be in one place at once. Your agents don't have that problem. Copilot agents in VS Code run three ways — local for interactive pairing, background for isolated tasks in worktrees, and cloud for autonomous long-running jobs. See when to reach for each, how to hand off between them, and manage parallel sessions from a single view. Custom agents, subagent orchestration, and delegating a local plan to background or cloud execution are all on the table.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:50 - 22:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711903001rZpY_20260512022411450.avif","speakerIds":["1768245294609build26-1774303544298001acRa","1768245294609build26-1774285000548001pBWh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.852Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711903001rZpY_20260512022411450.avif","ogTitle":"Fire, forget, sleep. VS Code agents that run without you","ogDescription":"You can only be in one place at once. Your agents don't have that problem. Copilot agents in VS Code run three ways — local for interactive pairing, background for isolated tasks in worktrees, and cloud for autonomous long-running jobs. See when to reach for each, how to hand off between them, and manage parallel sessions from a single view. Custom agents, subagent orchestration, and delegating a local plan to background or cloud execution are all on the table.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000548001pBWh","RegistrantKey":"1768245294609build26-1774285000548001pBWh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774303544298001acRa","RegistrantKey":"1768245294609build26-1774303544298001acRa"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"e8dba580-97a2-462f-a134-56137c462235","startDateTime":"2026-06-03T00:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maanav Dalal, Meng Tang, Alyanna Castillo","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018720412001rxAl","langLocale":"en-US","sessionId":"1774018720412001rxAl","sessionInstanceId":"1774018720412001rxAl","sessionCode":"LTG435","sessionCodeNormalized":"EN-US-LTG435","title":"Foundry Local in Action: AI on Windows Devices","sortTitle":"foundry local in action: ai on windows devices","description":"Join us for a session on getting started with Foundry Local on Windows, where we'll cover best practices and tips to help you make the most of using it on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:10 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720412001rxAl_20260512023111602.avif","speakerIds":["1768245294609build26-1774887302528001Bhs4","1768245294609build26-1774383196037001nAji","1768245294609build26-1772650125117001XJap"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.632Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720412001rxAl_20260512023111602.avif","ogTitle":"Foundry Local in Action: AI on Windows Devices","ogDescription":"Join us for a session on getting started with Foundry Local on Windows, where we'll cover best practices and tips to help you make the most of using it on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774383196037001nAji","RegistrantKey":"1768245294609build26-1774383196037001nAji"},{"SpeakerType":"Content Lead, Session Owner","SpeakerId":"1768245294609build26-1772650125117001XJap","RegistrantKey":"1768245294609build26-1772650125117001XJap"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774887302528001Bhs4","RegistrantKey":"1768245294609build26-1774887302528001Bhs4"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"}]},{"@search.score":1.0,"documentId":"e1d46991-0282-4625-817d-921ccd0ffd08","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Julio Arruda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293046515001L7PE","langLocale":"en-US","sessionId":"1776293046515001L7PE","sessionInstanceId":"1776293046515001L7PE","sessionCode":"LTG456","sessionCodeNormalized":"EN-US-LTG456","title":"From COBOL to AI: Using Azure AI to Modernize What You Can’t Observe","sortTitle":"from cobol to ai: using azure ai to modernize what you can’t observe","description":"Legacy systems hold decades of critical business logic — but they’re opaque, hard to evolve, and nearly impossible to instrument properly. In this session, you’ll see how uses AI to do what used to take years: intelligently discover your application’s business domains and dependencies, extract the rules buried inside COBOL programs, and generate a modern, decoupled application ready for the cloud-native world.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:30 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293046515001L7PE_20260512030136596.avif","speakerIds":["1768245294609build26-1762906426121001M0Ri"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.888Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293046515001L7PE_20260512030136596.avif","ogTitle":"From COBOL to AI: Using Azure AI to Modernize What You Can’t Observe","ogDescription":"Legacy systems hold decades of critical business logic — but they’re opaque, hard to evolve, and nearly impossible to instrument properly. In this session, you’ll see how uses AI to do what used to take years: intelligently discover your application’s business domains and dependencies, extract the rules buried inside COBOL programs, and generate a modern, decoupled application ready for the cloud-native world.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762906426121001M0Ri","RegistrantKey":"1768245294609build26-1762906426121001M0Ri"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"798b3e9f-803f-43b9-9b61-73f57c4a2e15","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonah Andersson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779814327211001lX7R","langLocale":"en-US","sessionId":"1779814327211001lX7R","sessionInstanceId":"1779814327211001lX7R","sessionCode":"LTG455-R1","sessionCodeNormalized":"EN-US-LTG455-R1","title":"From Code to Agents: Build Production MCP Servers on Azure Functions","sortTitle":"from code to agents: build production mcp servers on azure functions","description":"Agents are transforming how we build software but taking them from prototype to production remains the biggest challenge enterprise teams face. AI Agents need to call real tools- and today that means fragile, custom integrations that break when anything changes. MCP gives agents a standard way to discover and invoke tools at runtime. See real-world scenarios such as a DevOps CI/CD automation system where agents identify detect deployment issues, diagnose, and create rollback PRs in Azure DevOps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:00 - 00:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814327211001lX7R_20260527112819638.avif","speakerIds":["1768245294609build26-1776293017373001FIeV"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.239Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814327211001lX7R_20260527112819638.avif","ogTitle":"From Code to Agents: Build Production MCP Servers on Azure Functions","ogDescription":"Agents are transforming how we build software but taking them from prototype to production remains the biggest challenge enterprise teams face. AI Agents need to call real tools- and today that means fragile, custom integrations that break when anything changes. MCP gives agents a standard way to discover and invoke tools at runtime. See real-world scenarios such as a DevOps CI/CD automation system where agents identify detect deployment issues, diagnose, and create rollback PRs in Azure DevOps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293017373001FIeV","RegistrantKey":"1768245294609build26-1776293017373001FIeV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"141b2c29-e22c-4ac6-9b73-493761d8d942","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jonah Andersson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776292966476001xEB0","langLocale":"en-US","sessionId":"1776292966476001xEB0","sessionInstanceId":"1776292966476001xEB0","sessionCode":"LTG455","sessionCodeNormalized":"EN-US-LTG455","title":"From Code to Agents: Build Production MCP Servers on Azure Functions","sortTitle":"from code to agents: build production mcp servers on azure functions","description":"Agents are transforming how we build software but taking them from prototype to production remains the biggest challenge enterprise teams face. AI Agents need to call real tools- and today that means fragile, custom integrations that break when anything changes. MCP gives agents a standard way to discover and invoke tools at runtime. See real-world scenarios such as a DevOps CI/CD automation system where agents identify detect deployment issues, diagnose, and create rollback PRs in Azure DevOps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:10 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292966476001xEB0_20260512030127481.avif","speakerIds":["1768245294609build26-1776293017373001FIeV"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.889Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776292966476001xEB0_20260512030127481.avif","ogTitle":"From Code to Agents: Build Production MCP Servers on Azure Functions","ogDescription":"Agents are transforming how we build software but taking them from prototype to production remains the biggest challenge enterprise teams face. AI Agents need to call real tools- and today that means fragile, custom integrations that break when anything changes. MCP gives agents a standard way to discover and invoke tools at runtime. See real-world scenarios such as a DevOps CI/CD automation system where agents identify detect deployment issues, diagnose, and create rollback PRs in Azure DevOps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293017373001FIeV","RegistrantKey":"1768245294609build26-1776293017373001FIeV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"d8808647-b0fb-4350-99ab-9879639d5a26","startDateTime":"2026-06-02T19:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayca Bas","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703337001rwWK","langLocale":"en-US","sessionId":"1774018703337001rwWK","sessionInstanceId":"1774018703337001rwWK","sessionCode":"LTG424","sessionCodeNormalized":"EN-US-LTG424","title":"From fragile demos to production agents with Foundry Toolkit","sortTitle":"from fragile demos to production agents with foundry toolkit","description":"Building AI agents is easy to demo, but hard to iterate, debug, and ship with confidence. Prompts, logic, and deployment often live in different tools, making it difficult to understand behavior or catch regressions. This talk shows how to design, build, and refine agents in VS Code with Foundry Toolkit and Copilot, evaluate and trace their behavior, and deploy to Microsoft Foundry—so prototypes can evolve into production without rework.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:20 - 19:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703337001rwWK_20260518110517888.avif","speakerIds":["1768245294609build26-1762250455017001UQP1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.459Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703337001rwWK_20260518110517888.avif","ogTitle":"From fragile demos to production agents with Foundry Toolkit","ogDescription":"Building AI agents is easy to demo, but hard to iterate, debug, and ship with confidence. Prompts, logic, and deployment often live in different tools, making it difficult to understand behavior or catch regressions. This talk shows how to design, build, and refine agents in VS Code with Foundry Toolkit and Copilot, evaluate and trace their behavior, and deploy to Microsoft Foundry—so prototypes can evolve into production without rework.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762250455017001UQP1","RegistrantKey":"1768245294609build26-1762250455017001UQP1"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Tracing","logicalValue":"Tracing"},{"displayValue":"Evaluations","logicalValue":"Evaluations"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"78e57e33-19b0-4f2e-9836-b542bcff7915","startDateTime":"2026-06-03T23:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Andy Mutz","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779487666889001AwRY","langLocale":"en-US","sessionId":"1779487666889001AwRY","sessionInstanceId":"1779487666889001AwRY","sessionCode":"LTG407","sessionCodeNormalized":"EN-US-LTG407","title":"From Fragmented Data to Agentic Intelligence at Enterprise Scale—Powered by Fabric + Foundry","sortTitle":"from fragmented data to agentic intelligence at enterprise scale—powered by fabric + foundry","description":"In this lightning talk, we demonstrate how to transform fragmented customer experience data into production-ready AI agents—at enterprise scale. Starting with customer feedback as the source of truth, we convert raw signals into structured intelligence on Fabric—powering real-time, agent-driven insights and actions directly in the flow of work. Built for production, with Entra security, DLP, guardrails, and end-to-end telemetry seamlessly integrated.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:20 - 23:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779487666889001AwRY_20260527113146962.avif","speakerIds":["1768245294609build26-1728676736409001oIow"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.237Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779487666889001AwRY_20260527113146962.avif","ogTitle":"From Fragmented Data to Agentic Intelligence at Enterprise Scale—Powered by Fabric + Foundry","ogDescription":"In this lightning talk, we demonstrate how to transform fragmented customer experience data into production-ready AI agents—at enterprise scale. Starting with customer feedback as the source of truth, we convert raw signals into structured intelligence on Fabric—powering real-time, agent-driven insights and actions directly in the flow of work. Built for production, with Entra security, DLP, guardrails, and end-to-end telemetry seamlessly integrated.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1728676736409001oIow","RegistrantKey":"1768245294609build26-1728676736409001oIow"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"1638554b-3328-415a-a09e-8fa1dcf3b2d3","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Renato Romao de Souza","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779814187386001geyx","langLocale":"en-US","sessionId":"1779814187386001geyx","sessionInstanceId":"1779814187386001geyx","sessionCode":"LTG450-R1","sessionCodeNormalized":"EN-US-LTG450-R1","title":"From Idea to Agent in Minutes: Copilot Studio + Claude Code","sortTitle":"from idea to agent in minutes: copilot studio + claude code","description":"Building AI agents usually involves writing instructions, defining skills, structuring knowledge sources, and configuring actions. In this session, we’ll explore a new development pattern where AI assists the creation of other AI agents by combining Copilot Studio with Claude Code, instead of manually configuring every element inside the Copilot Studio interface.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:00 - 00:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814187386001geyx_20260527112757349.avif","speakerIds":["1768245294609build26-1756855668236001z806"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.239Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779814187386001geyx_20260527112757349.avif","ogTitle":"From Idea to Agent in Minutes: Copilot Studio + Claude Code","ogDescription":"Building AI agents usually involves writing instructions, defining skills, structuring knowledge sources, and configuring actions. In this session, we’ll explore a new development pattern where AI assists the creation of other AI agents by combining Copilot Studio with Claude Code, instead of manually configuring every element inside the Copilot Studio interface.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756855668236001z806","RegistrantKey":"1768245294609build26-1756855668236001z806"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"c094eaff-dcba-418d-a11f-fd2699ebdded","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Renato Romao de Souza","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290805423001iw5c","langLocale":"en-US","sessionId":"1776290805423001iw5c","sessionInstanceId":"1776290805423001iw5c","sessionCode":"LTG450","sessionCodeNormalized":"EN-US-LTG450","title":"From Idea to Agent in Minutes: Copilot Studio + Claude Code","sortTitle":"from idea to agent in minutes: copilot studio + claude code","description":"Building AI agents usually involves writing instructions, defining skills, structuring knowledge sources, and configuring actions. In this session, we’ll explore a new development pattern where AI assists the creation of other AI agents by combining Copilot Studio with Claude Code, instead of manually configuring every element inside the Copilot Studio interface.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:30 - 17:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776290805423001iw5c_20260512025940909.avif","speakerIds":["1768245294609build26-1756855668236001z806"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.887Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776290805423001iw5c_20260512025940909.avif","ogTitle":"From Idea to Agent in Minutes: Copilot Studio + Claude Code","ogDescription":"Building AI agents usually involves writing instructions, defining skills, structuring knowledge sources, and configuring actions. In this session, we’ll explore a new development pattern where AI assists the creation of other AI agents by combining Copilot Studio with Claude Code, instead of manually configuring every element inside the Copilot Studio interface.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1756855668236001z806","RegistrantKey":"1768245294609build26-1756855668236001z806"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"a27d0de6-a5ae-47f0-98a5-25d049c76447","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ashish Garg, Karthik Vijayan, Morris Novello","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888798445001fBEr","langLocale":"en-US","sessionId":"1776888798445001fBEr","sessionInstanceId":"1776888798445001fBEr","sessionCode":"LTGSP492","sessionCodeNormalized":"EN-US-LTGSP492","title":"From kernel to agent: Build on-device AI with 2-bit models","sortTitle":"from kernel to agent: build on-device ai with 2-bit models","description":"Running AI models locally is key to reducing latency, cost, and power usage—but model size remains a constraint. This session explores the engineering behind Microsoft and Qualcomm's 2-bit quantization. Learn how extreme quantization preserves model quality while dramatically reducing memory footprint. Understand how to use Windows ML APIs to load, run, and orchestrate quantized models locally on an AI PC to develop agentic AI applications.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:30 - 21:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888798445001fBEr_20260512031431056.avif","speakerIds":["1768245294609build26-1776966792659001kgb1","1768245294609build26-1777588199349001unfM","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.386Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888798445001fBEr_20260512031431056.avif","ogTitle":"From kernel to agent: Build on-device AI with 2-bit models","ogDescription":"Running AI models locally is key to reducing latency, cost, and power usage—but model size remains a constraint. This session explores the engineering behind Microsoft and Qualcomm's 2-bit quantization. Learn how extreme quantization preserves model quality while dramatically reducing memory footprint. Understand how to use Windows ML APIs to load, run, and orchestrate quantized models locally on an AI PC to develop agentic AI applications.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776966792659001kgb1","RegistrantKey":"1768245294609build26-1776966792659001kgb1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777588199349001unfM","RegistrantKey":"1768245294609build26-1777588199349001unfM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"497888c7-bfb8-44c4-89c6-050683bf2c45","startDateTime":"2026-06-02T23:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nik Karpinsky","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018712237001rLex","langLocale":"en-US","sessionId":"1774018712237001rLex","sessionInstanceId":"1774018712237001rLex","sessionCode":"LTG406","sessionCodeNormalized":"EN-US-LTG406","title":"Get More Than Autocomplete: GitHub Copilot Workflows in Visual Studio","sortTitle":"get more than autocomplete: github copilot workflows in visual studio","description":"GitHub Copilot in Visual Studio goes far beyond code completion. In this lightning talk, see how developers use Copilot across the full inner loop — from understanding unfamiliar code, to generating tests, debugging failures, and reviewing changes with confidence. We’ll highlight practical workflows that help teams move faster without losing control, all inside Visual Studio.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:10 - 23:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018712237001rLex_20260512022430307.avif","speakerIds":["1768245294609build26-1774645705844001UesE"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.853Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018712237001rLex_20260512022430307.avif","ogTitle":"Get More Than Autocomplete: GitHub Copilot Workflows in Visual Studio","ogDescription":"GitHub Copilot in Visual Studio goes far beyond code completion. In this lightning talk, see how developers use Copilot across the full inner loop — from understanding unfamiliar code, to generating tests, debugging failures, and reviewing changes with confidence. We’ll highlight practical workflows that help teams move faster without losing control, all inside Visual Studio.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774645705844001UesE","RegistrantKey":"1768245294609build26-1774645705844001UesE"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"caaf7918-233f-4d30-841e-72318a93ee24","startDateTime":"2026-06-03T21:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alex Thissen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293477835001KbIy","langLocale":"en-US","sessionId":"1776293477835001KbIy","sessionInstanceId":"1776293477835001KbIy","sessionCode":"LTG462","sessionCodeNormalized":"EN-US-LTG462","title":"Hidden gems of Microsoft tools: Dev Proxy and Dev Tunnel","sortTitle":"hidden gems of microsoft tools: dev proxy and dev tunnel","description":"In this session, you’ll discover two lesser-known Microsoft tools: Dev Proxy and Dev Tunnel. Learn how Dev Proxy enables reality-driven development by simulating failures, mock responses and giving you control over real-world conditions. You’ll also see how Dev Tunnel lets you expose locally running services to the cloud, making it easy to prototype, test and collaborate across environments. Walk away with practical techniques to build resilient applications and unlock new development workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:40 - 21:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293477835001KbIy_20260512030256571.avif","speakerIds":["1768245294609build26-1772633942097001tO2F"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.889Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293477835001KbIy_20260512030256571.avif","ogTitle":"Hidden gems of Microsoft tools: Dev Proxy and Dev Tunnel","ogDescription":"In this session, you’ll discover two lesser-known Microsoft tools: Dev Proxy and Dev Tunnel. Learn how Dev Proxy enables reality-driven development by simulating failures, mock responses and giving you control over real-world conditions. You’ll also see how Dev Tunnel lets you expose locally running services to the cloud, making it easy to prototype, test and collaborate across environments. Walk away with practical techniques to build resilient applications and unlock new development workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772633942097001tO2F","RegistrantKey":"1768245294609build26-1772633942097001tO2F"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"29475896-831d-4e8a-a558-3fa256be0416","startDateTime":"2026-06-02T19:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Partha Srinivasan, Can Tuzla","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774543788276001lyfy","langLocale":"en-US","sessionId":"1774543788276001lyfy","sessionInstanceId":"1774543788276001lyfy","sessionCode":"LTGSP488","sessionCodeNormalized":"EN-US-LTGSP488","title":"Improve analytic models with Oracle Data Science Agent and Azure OpenAI","sortTitle":"improve analytic models with oracle data science agent and azure openai","description":"Learn how to turn simple conversational prompts into highly accurate predictions within a governed, developer-friendly workflow. This session explores building analytics workflows using Oracle Data Science Agent with Oracle Autonomous AI Database on Oracle AI Database@Azure. Gain practical tips on profiling data from Oracle Autonomous AI Database and Azure Blob Storage to uncover patterns, train models, and generate inference-ready SQL using OpenAI-powered Data Science Agent workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:20 - 19:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774543788276001lyfy_20260528130828198.avif","speakerIds":["1768245294609build26-1776389175315001yxb8","1768245294609build26-1778294924883001Yt7v"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.841Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Oracle"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774543788276001lyfy_20260528130828198.avif","ogTitle":"Improve analytic models with Oracle Data Science Agent and Azure OpenAI","ogDescription":"Learn how to turn simple conversational prompts into highly accurate predictions within a governed, developer-friendly workflow. This session explores building analytics workflows using Oracle Data Science Agent with Oracle Autonomous AI Database on Oracle AI Database@Azure. Gain practical tips on profiling data from Oracle Autonomous AI Database and Azure Blob Storage to uncover patterns, train models, and generate inference-ready SQL using OpenAI-powered Data Science Agent workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776389175315001yxb8","RegistrantKey":"1768245294609build26-1776389175315001yxb8"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778294924883001Yt7v","RegistrantKey":"1768245294609build26-1778294924883001Yt7v"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Integration","logicalValue":"App Integration"}]},{"@search.score":1.0,"documentId":"f878316c-90b9-4ee3-9d55-e19722c67c14","startDateTime":"2026-06-03T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Julio Arruda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293541060001kUtB","langLocale":"en-US","sessionId":"1776293541060001kUtB","sessionInstanceId":"1776293541060001kUtB","sessionCode":"LTG463","sessionCodeNormalized":"EN-US-LTG463","title":"In AI, Specifications Are Not Contracts — They’re Hypotheses","sortTitle":"in ai, specifications are not contracts — they’re hypotheses","description":"As generative AI becomes part of modern software development, teams are increasingly relying on specifications to guide the behavior of AI agents, copilots, and intelligent workflows powered by tools like GitHub Copilot. But there’s a subtle trap: treating specifications the same way we did in traditional software engineering.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:00 - 23:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293541060001kUtB_20260512030305994.avif","speakerIds":["1768245294609build26-1762906426121001M0Ri"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.89Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293541060001kUtB_20260512030305994.avif","ogTitle":"In AI, Specifications Are Not Contracts — They’re Hypotheses","ogDescription":"As generative AI becomes part of modern software development, teams are increasingly relying on specifications to guide the behavior of AI agents, copilots, and intelligent workflows powered by tools like GitHub Copilot. But there’s a subtle trap: treating specifications the same way we did in traditional software engineering.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1762906426121001M0Ri","RegistrantKey":"1768245294609build26-1762906426121001M0Ri"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"8577ea62-6f02-4030-ac60-e5126b37a5dc","startDateTime":"2026-06-03T17:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Michel Hubert","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293115454001KHxE","langLocale":"en-US","sessionId":"1776293115454001KHxE","sessionInstanceId":"1776293115454001KHxE","sessionCode":"LTG457","sessionCodeNormalized":"EN-US-LTG457","title":"Infrastructure for AI: Deploying a Production-Ready Foundation","sortTitle":"infrastructure for ai: deploying a production-ready foundation","description":"AI workloads do not just change applications; they change what infrastructure needs to deliver. This session uses a practical demo to show how to provision and prepare an environment for AI-enabled applications, with attention to scalability, observability, resilience, and operational readiness. The aim is to make infrastructure strategy visible through a concrete example.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:10 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293115454001KHxE_20260512030150838.avif","speakerIds":["1768245294609build26-1757452741184001uVMn"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.89Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293115454001KHxE_20260512030150838.avif","ogTitle":"Infrastructure for AI: Deploying a Production-Ready Foundation","ogDescription":"AI workloads do not just change applications; they change what infrastructure needs to deliver. This session uses a practical demo to show how to provision and prepare an environment for AI-enabled applications, with attention to scalability, observability, resilience, and operational readiness. The aim is to make infrastructure strategy visible through a concrete example.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1757452741184001uVMn","RegistrantKey":"1768245294609build26-1757452741184001uVMn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"7117b9c6-20f4-4412-ac01-5cbe2c514020","startDateTime":"2026-06-03T00:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sharmila Chockalingam, Caroline Matthews","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018715313001r3eu","langLocale":"en-US","sessionId":"1774018715313001r3eu","sessionInstanceId":"1774018715313001r3eu","sessionCode":"LTG416","sessionCodeNormalized":"EN-US-LTG416","title":"Long-horizon coding agents with Claude in Microsoft Foundry","sortTitle":"long-horizon coding agents with claude in microsoft foundry","description":"AI-driven coding requires more than code generation, agents must know when work is complete and whether the result is actually correct. This session explores the architecture behind long-horizon coding agents with Claude in Microsoft Foundry, where planners break prompts into tasks, generators build solutions, and evaluators validate results in a real browser. Includes a live walkthrough from a single prompt to working software.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:10 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715313001r3eu_20260529120925847.avif","speakerIds":["1768245294609build26-1752517267116001lWQS","1768245294609build26-1778020065215001t4NN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.855Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715313001r3eu_20260529120925847.avif","ogTitle":"Long-horizon coding agents with Claude in Microsoft Foundry","ogDescription":"AI-driven coding requires more than code generation, agents must know when work is complete and whether the result is actually correct. This session explores the architecture behind long-horizon coding agents with Claude in Microsoft Foundry, where planners break prompts into tasks, generators build solutions, and evaluators validate results in a real browser. Includes a live walkthrough from a single prompt to working software.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1752517267116001lWQS","RegistrantKey":"1768245294609build26-1752517267116001lWQS"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778020065215001t4NN","RegistrantKey":"1768245294609build26-1778020065215001t4NN"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"6ef06e47-e325-436a-98ca-d8ad985b367e","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeff Blankenburg","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392407043001iVGz","langLocale":"en-US","sessionId":"1774392407043001iVGz","sessionInstanceId":"1774392407043001iVGz","sessionCode":"LTGSP484","sessionCodeNormalized":"EN-US-LTGSP484","title":"Make sure you understand the code AI is writing","sortTitle":"make sure you understand the code ai is writing","description":"Copilot ships code fast. But neither you nor your agents understand what it does in production. That gap is growing. The old loop still looks like:\ndeploy → dashboards → logs → redeploy\n\nIt wasn’t built for AI-paced development. In this lightning talk, we break a real service in VS Code, then fix it using Copilot agent mode grounded in Dynatrace MCP. See runtime context flow into your editor so you and your agents can work in context and fix issues at the speed code is created.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:30 - 21:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392407043001iVGz_20260512023629044.avif","speakerIds":["1768245294609build26-1775666678371001JTOl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.637Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Dynatrace"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392407043001iVGz_20260512023629044.avif","ogTitle":"Make sure you understand the code AI is writing","ogDescription":"Copilot ships code fast. But neither you nor your agents understand what it does in production. That gap is growing. The old loop still looks like:\ndeploy → dashboards → logs → redeploy\n\nIt wasn’t built for AI-paced development. In this lightning talk, we break a real service in VS Code, then fix it using Copilot agent mode grounded in Dynatrace MCP. See runtime context flow into your editor so you and your agents can work in context and fix issues at the speed code is created.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775666678371001JTOl","RegistrantKey":"1768245294609build26-1775666678371001JTOl"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Outages","logicalValue":"Outages"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Monitor","logicalValue":"Monitor"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"4296e4f1-a575-4e69-820f-dc592d97b80c","startDateTime":"2026-06-03T19:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Tyler Leonhardt, Connor Peet","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711742001r2Eh","langLocale":"en-US","sessionId":"1774018711742001r2Eh","sessionInstanceId":"1774018711742001r2Eh","sessionCode":"LTG403","sessionCodeNormalized":"EN-US-LTG403","title":"MCP does way more than you think","sortTitle":"mcp does way more than you think","description":"You've connected an MCP server or two. But MCP in GitHub Copilot goes a lot further than tool lookups — and it works across the editor, CLI, and cloud. This session demos the range: connecting Copilot to databases, APIs, and internal systems wherever you work, plus a first look at MCP Apps, where tool calls return real UI like dashboards and forms right in chat. All live, all demos, no slides.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:30 - 19:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711742001r2Eh_20260512022402539.avif","speakerIds":["1768245294609build26-1774285000394001pcWF","1768245294609build26-1774285001351001pygt"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.79Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711742001r2Eh_20260512022402539.avif","ogTitle":"MCP does way more than you think","ogDescription":"You've connected an MCP server or two. But MCP in GitHub Copilot goes a lot further than tool lookups — and it works across the editor, CLI, and cloud. This session demos the range: connecting Copilot to databases, APIs, and internal systems wherever you work, plus a first look at MCP Apps, where tool calls return real UI like dashboards and forms right in chat. All live, all demos, no slides.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000394001pcWF","RegistrantKey":"1768245294609build26-1774285000394001pcWF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285001351001pygt","RegistrantKey":"1768245294609build26-1774285001351001pygt"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"4d079f3d-628f-427c-b111-81712805fad1","startDateTime":"2026-06-04T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Tyler Leonhardt, Connor Peet","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778263096083001RXQN","langLocale":"en-US","sessionId":"1778263096083001RXQN","sessionInstanceId":"1778263096083001RXQN","sessionCode":"LTG403-R1","sessionCodeNormalized":"EN-US-LTG403-R1","title":"MCP does way more than you think (Repeat)","sortTitle":"mcp does way more than you think (repeat)","description":"You've connected an MCP server or two. But MCP in GitHub Copilot goes a lot further than tool lookups — and it works across the editor, CLI, and cloud. This session demos the range: connecting Copilot to databases, APIs, and internal systems wherever you work, plus a first look at MCP Apps, where tool calls return real UI like dashboards and forms right in chat. All live, all demos, no slides.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:00 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:25:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263096083001RXQN_20260512031944204.avif","speakerIds":["1768245294609build26-1774285000394001pcWF","1768245294609build26-1774285001351001pygt"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.775Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263096083001RXQN_20260512031944204.avif","ogTitle":"MCP does way more than you think (Repeat)","ogDescription":"You've connected an MCP server or two. But MCP in GitHub Copilot goes a lot further than tool lookups — and it works across the editor, CLI, and cloud. This session demos the range: connecting Copilot to databases, APIs, and internal systems wherever you work, plus a first look at MCP Apps, where tool calls return real UI like dashboards and forms right in chat. All live, all demos, no slides.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater B","logicalValue":"Gateway Pavilion, Level 2, Theater B"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000394001pcWF","RegistrantKey":"1768245294609build26-1774285000394001pcWF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285001351001pygt","RegistrantKey":"1768245294609build26-1774285001351001pygt"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"83865bbc-ad2d-4758-a89a-32c2b5147c23","startDateTime":"2026-06-03T21:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Graeme Malcolm, Aaron Stewart","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774050846671001pphO","langLocale":"en-US","sessionId":"1774050846671001pphO","sessionInstanceId":"1774050846671001pphO","sessionCode":"LTG440","sessionCodeNormalized":"EN-US-LTG440","title":"Microsoft Credentials for Developers: See what’s next","sortTitle":"microsoft credentials for developers: see what’s next","description":"AI is changing how code gets written and shipped. This session provides an overview of Microsoft Credentials for developers and explores new ways to recognize real product proficiency as it shows up in everyday work with GitHub Copilot. Learn how the evolving credentials portfolio validates practical execution across real workflows, so you can prove what you can build, not just what you know.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:20 - 21:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050846671001pphO_20260512023249292.avif","speakerIds":["1768245294609build26-1775674794229001Nd1V","1768245294609build26-1776368652182001X6Mh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.633Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050846671001pphO_20260512023249292.avif","ogTitle":"Microsoft Credentials for Developers: See what’s next","ogDescription":"AI is changing how code gets written and shipped. This session provides an overview of Microsoft Credentials for developers and explores new ways to recognize real product proficiency as it shows up in everyday work with GitHub Copilot. Learn how the evolving credentials portfolio validates practical execution across real workflows, so you can prove what you can build, not just what you know.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776368652182001X6Mh","RegistrantKey":"1768245294609build26-1776368652182001X6Mh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775674794229001Nd1V","RegistrantKey":"1768245294609build26-1775674794229001Nd1V"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Skills","logicalValue":"Skills"}]},{"@search.score":1.0,"documentId":"4bb48e6f-380e-4738-820f-58151f81230c","startDateTime":"2026-06-03T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Felipe Ospina","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703013001rFaO","langLocale":"en-US","sessionId":"1774018703013001rFaO","sessionInstanceId":"1774018703013001rFaO","sessionCode":"LTG422","sessionCodeNormalized":"EN-US-LTG422","title":"Monetize apps and agents with Microsoft Marketplace","sortTitle":"monetize apps and agents with microsoft marketplace","description":"Learn how to build, publish, and monetize apps and agents with Microsoft Marketplace to reach millions of customers and drive scalable growth. Getting published\u00A0on Marketplace\u00A0surfaces your solutions in the Marketplace storefront and\u00A0Microsoft products, such as\u00A0Microsoft 365 Copilot and Azure, to reach customers directly in the flow of work. This lightning talk covers the start-to-finish publishing process and highlights\u00A0benefits and incentives available from Microsoft\u00A0for software developers.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:30 - 00:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703013001rFaO_20260512021403083.avif","speakerIds":["1768245294609build26-1763598207239001O4Hn"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.458Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703013001rFaO_20260512021403083.avif","ogTitle":"Monetize apps and agents with Microsoft Marketplace","ogDescription":"Learn how to build, publish, and monetize apps and agents with Microsoft Marketplace to reach millions of customers and drive scalable growth. Getting published\u00A0on Marketplace\u00A0surfaces your solutions in the Marketplace storefront and\u00A0Microsoft products, such as\u00A0Microsoft 365 Copilot and Azure, to reach customers directly in the flow of work. This lightning talk covers the start-to-finish publishing process and highlights\u00A0benefits and incentives available from Microsoft\u00A0for software developers.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1763598207239001O4Hn","RegistrantKey":"1768245294609build26-1763598207239001O4Hn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"ISV","logicalValue":"ISV"}]},{"@search.score":1.0,"documentId":"3d82f797-7ee1-4c4c-929f-d4382603f1ee","startDateTime":"2026-06-03T23:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kate Yoak","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392583909001syyk","langLocale":"en-US","sessionId":"1774392583909001syyk","sessionInstanceId":"1774392583909001syyk","sessionCode":"LTGSP486","sessionCodeNormalized":"EN-US-LTGSP486","title":"Monitor and optimize GenAI workloads on Azure","sortTitle":"monitor and optimize genai workloads on azure","description":"Monitoring GenAI applications requires more than traditional telemetry—latency, cost, and model quality all matter.\nLearn how to monitor and optimize large language model workloads on Azure using Azure OpenAI, App Services, and Datadog. Explore real-world patterns for tracking latency, evaluating model quality, and measuring usage at scale.\nWalk away with practical strategies to keep GenAI applications reliable, cost-efficient, and performant in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:20 - 23:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392583909001syyk_20260512023651489.avif","speakerIds":["1768245294609build26-1776716407683001IPXm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.639Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Datadog"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392583909001syyk_20260512023651489.avif","ogTitle":"Monitor and optimize GenAI workloads on Azure","ogDescription":"Monitoring GenAI applications requires more than traditional telemetry—latency, cost, and model quality all matter.\nLearn how to monitor and optimize large language model workloads on Azure using Azure OpenAI, App Services, and Datadog. Explore real-world patterns for tracking latency, evaluating model quality, and measuring usage at scale.\nWalk away with practical strategies to keep GenAI applications reliable, cost-efficient, and performant in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776716407683001IPXm","RegistrantKey":"1768245294609build26-1776716407683001IPXm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Openclaw","logicalValue":"Openclaw"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"ca6d3e04-0051-4dc5-aae8-0ef602b1f1de","startDateTime":"2026-06-03T20:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Masaki Kumamoto, Sandeep Panda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392472149001egUQ","langLocale":"en-US","sessionId":"1774392472149001egUQ","sessionInstanceId":"1774392472149001egUQ","sessionCode":"LTGSP485","sessionCodeNormalized":"EN-US-LTGSP485","title":"Move AI from pilot to production with  UiPath + Microsoft","sortTitle":"move ai from pilot to production with  uipath + microsoft","description":"Many agentic AI solutions stall at pilot stage due to gaps in orchestration and governance. In this lightning talk, see a joint enterprise architecture for scaling agentic automation using Microsoft and UiPath. See how to connect agents, Copilot, and workflows into a governed architecture, and apply this approach to move from experimentation to production\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:40 - 20:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392472149001egUQ_20260526104856594.avif","speakerIds":["1768245294609build26-1778165745431001j8OM","1768245294609build26-1765885779904001cs4C"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.638Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["UiPath"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392472149001egUQ_20260526104856594.avif","ogTitle":"Move AI from pilot to production with  UiPath + Microsoft","ogDescription":"Many agentic AI solutions stall at pilot stage due to gaps in orchestration and governance. In this lightning talk, see a joint enterprise architecture for scaling agentic automation using Microsoft and UiPath. See how to connect agents, Copilot, and workflows into a governed architecture, and apply this approach to move from experimentation to production\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778165745431001j8OM","RegistrantKey":"1768245294609build26-1778165745431001j8OM"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765885779904001cs4C","RegistrantKey":"1768245294609build26-1765885779904001cs4C"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Copilot Studio","logicalValue":"Copilot Studio"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"}]},{"@search.score":1.0,"documentId":"88b0742f-c9e5-4b7d-8592-aa2e7928236b","startDateTime":"2026-06-03T22:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gaurav Goel, Morris Novello","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888715474001ItwE","langLocale":"en-US","sessionId":"1776888715474001ItwE","sessionInstanceId":"1776888715474001ItwE","sessionCode":"LTGSP497","sessionCodeNormalized":"EN-US-LTGSP497","title":"Port x86 Windows apps to ARM on Snapdragon with AI","sortTitle":"port x86 windows apps to arm on snapdragon with ai","description":"Learn how to optimize and port Windows apps for ARM on Snapdragon with minimal code changes. This session covers the developer toolchain, including compilers, debuggers, profilers, and GitHub Actions, along with current ecosystem support. See where ARM performance gains are real today, and how agentic AI can help convert and validate x86 applications for speed, compatibility, and scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:20 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888715474001ItwE_20260512031412419.avif","speakerIds":["1768245294609build26-1777404623082001WcZo","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.385Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888715474001ItwE_20260512031412419.avif","ogTitle":"Port x86 Windows apps to ARM on Snapdragon with AI","ogDescription":"Learn how to optimize and port Windows apps for ARM on Snapdragon with minimal code changes. This session covers the developer toolchain, including compilers, debuggers, profilers, and GitHub Actions, along with current ecosystem support. See where ARM performance gains are real today, and how agentic AI can help convert and validate x86 applications for speed, compatibility, and scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777404623082001WcZo","RegistrantKey":"1768245294609build26-1777404623082001WcZo"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Software Development Company","logicalValue":"Software Development Company"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"Win32 Developers","logicalValue":"Win32 Developers"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"App Integration","logicalValue":"App Integration"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Windows applications","logicalValue":"Windows applications"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"},{"displayValue":"Development pipeline","logicalValue":"Development pipeline"}]},{"@search.score":1.0,"documentId":"0608f26c-a934-4956-84c9-620dd7a717e1","startDateTime":"2026-06-03T18:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Fatih Akin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774051373338001BTGG","langLocale":"en-US","sessionId":"1774051373338001BTGG","sessionInstanceId":"1774051373338001BTGG","sessionCode":"LTG468","sessionCodeNormalized":"EN-US-LTG468","title":"prompts.chat: Building an Open Prompt Ecosystem for the Age of Agents","sortTitle":"prompts.chat: building an open prompt ecosystem for the age of agents","description":"Agents are getting more capable, but prompts still matter. The difference is prompts are no longer one-off inputs. They are becoming reusable building blocks for workflows, tools, and shared AI systems. I’ll share the story of prompts.chat and how it evolved from an open-source prompt library into a broader ecosystem, why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:10 - 18:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051373338001BTGG_20260512023335474.avif","speakerIds":["1768245294609build26-1776361485371001HMvG"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.634Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051373338001BTGG_20260512023335474.avif","ogTitle":"prompts.chat: Building an Open Prompt Ecosystem for the Age of Agents","ogDescription":"Agents are getting more capable, but prompts still matter. The difference is prompts are no longer one-off inputs. They are becoming reusable building blocks for workflows, tools, and shared AI systems. I’ll share the story of prompts.chat and how it evolved from an open-source prompt library into a broader ecosystem, why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776361485371001HMvG","RegistrantKey":"1768245294609build26-1776361485371001HMvG"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"OSS CI/CD Libraries","logicalValue":"OSS CI/CD Libraries"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"431ca3d7-4cd7-4855-b3e2-5ac224045aeb","startDateTime":"2026-06-02T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Lee Stott","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018715626001rYEl","langLocale":"en-US","sessionId":"1774018715626001rYEl","sessionInstanceId":"1774018715626001rYEl","sessionCode":"LTG417","sessionCodeNormalized":"EN-US-LTG417","title":"Quick wins for AI model selection with Microsoft Foundry","sortTitle":"quick wins for ai model selection with microsoft foundry","description":"With hundreds of AI models available, choosing and integrating the right one can be daunting. This lightning talk shares practical tips for evaluating, benchmarking, and adopting models using Microsoft Foundry’s streamlined tools. Learn how IDE integration and validation workflows help you start projects quickly and avoid common pitfalls.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:30 - 22:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715626001rYEl_20260512022654675.avif","speakerIds":["1768245294609build26-1758764115124001fdmg"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.855Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715626001rYEl_20260512022654675.avif","ogTitle":"Quick wins for AI model selection with Microsoft Foundry","ogDescription":"With hundreds of AI models available, choosing and integrating the right one can be daunting. This lightning talk shares practical tips for evaluating, benchmarking, and adopting models using Microsoft Foundry’s streamlined tools. Learn how IDE integration and validation workflows help you start projects quickly and avoid common pitfalls.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1758764115124001fdmg","RegistrantKey":"1768245294609build26-1758764115124001fdmg"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"d7e5dfa2-a742-4cd8-8f72-be78b681b9e4","startDateTime":"2026-06-03T16:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Surya Kukkapalli","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774050911684001kOEr","langLocale":"en-US","sessionId":"1774050911684001kOEr","sessionInstanceId":"1774050911684001kOEr","sessionCode":"LTG442","sessionCodeNormalized":"EN-US-LTG442","title":"Revora Health: Imagine Cup Finalist Scaling AI Recovery","sortTitle":"revora health: imagine cup finalist scaling ai recovery","description":"Scale is the biggest challenge in physical therapy. Join Revora Health for a deep dive into an Azure‑powered platform that connects patients to licensed providers in minutes, not months. Built on Azure Health Data Services and AI‑driven computer vision, Revora Health delivers real‑time, multimodal guidance that improves adherence while reducing provider burnout. HIPAA‑compliant Azure AI enables continuous, accessible recovery without removing clinicians from the loop.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:50 - 17:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050911684001kOEr_20260512023311922.avif","speakerIds":["1768245294609build26-1775834985858001az7i"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.634Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774050911684001kOEr_20260512023311922.avif","ogTitle":"Revora Health: Imagine Cup Finalist Scaling AI Recovery","ogDescription":"Scale is the biggest challenge in physical therapy. Join Revora Health for a deep dive into an Azure‑powered platform that connects patients to licensed providers in minutes, not months. Built on Azure Health Data Services and AI‑driven computer vision, Revora Health delivers real‑time, multimodal guidance that improves adherence while reducing provider burnout. HIPAA‑compliant Azure AI enables continuous, accessible recovery without removing clinicians from the loop.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775834985858001az7i","RegistrantKey":"1768245294609build26-1775834985858001az7i"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Azure ML","logicalValue":"Azure ML"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Personalization","logicalValue":"Personalization"}]},{"@search.score":1.0,"documentId":"03715acf-9d15-45af-b04b-b7918ffee214","startDateTime":"2026-06-02T22:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mehrnoosh Sameki, Laura Wolf","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018703795001rM81","langLocale":"en-US","sessionId":"1774018703795001rM81","sessionInstanceId":"1774018703795001rM81","sessionCode":"LTG427","sessionCodeNormalized":"EN-US-LTG427","title":"Rubric-Based Evaluation for Real World AI Systems","sortTitle":"rubric-based evaluation for real world ai systems","description":"AI teams need to prove their systems behave responsibly, but turning policies, safety requirements, or product expectations into evaluations is still slow and manual. This talk introduces a way to translate natural-language requirements into testable artifacts: concrete behaviors, targeted test cases, evaluation runs, and evidence of where the system aligns or fails. It helps teams move from “we have a requirement” to “we can measure whether the system follows it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:50 - 23:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703795001rM81_20260512021450485.avif","speakerIds":["1768245294609build26-1775085874620001ADSc","1768245294609build26-1767984116727001aLg7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.459Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018703795001rM81_20260512021450485.avif","ogTitle":"Rubric-Based Evaluation for Real World AI Systems","ogDescription":"AI teams need to prove their systems behave responsibly, but turning policies, safety requirements, or product expectations into evaluations is still slow and manual. This talk introduces a way to translate natural-language requirements into testable artifacts: concrete behaviors, targeted test cases, evaluation runs, and evidence of where the system aligns or fails. It helps teams move from “we have a requirement” to “we can measure whether the system follows it.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Content Lead, Session Owner","SpeakerId":"1768245294609build26-1767984116727001aLg7","RegistrantKey":"1768245294609build26-1767984116727001aLg7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775085874620001ADSc","RegistrantKey":"1768245294609build26-1775085874620001ADSc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Evaluations","logicalValue":"Evaluations"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"261cb18b-b8d9-4331-a62a-4c8a2bab3c87","startDateTime":"2026-06-03T18:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Manuj Sabharwal, Gokul Tonpe, Morris Novello","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776888767709001tN4X","langLocale":"en-US","sessionId":"1776888767709001tN4X","sessionInstanceId":"1776888767709001tN4X","sessionCode":"LTGSP493","sessionCodeNormalized":"EN-US-LTGSP493","title":"Run AI coding agents on-device with NPUs","sortTitle":"run ai coding agents on-device with npus","description":"AI-powered coding agents are transforming developer workflows, but most run in the cloud, adding latency, cost, and privacy concerns. What if your coding agent ran entirely on your laptop’s NPU? In this session, you'll learn how Visual Studio AI Toolkit and Qualcomm’s Hexagon NPU enable fully on device agentic coding. Using optimized SLMs, you'll experience real time code generation with no cloud calls, delivering high performance per watt, longer battery life, and responsive developer UX.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:10 - 18:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888767709001tN4X_20260512031421430.avif","speakerIds":["1768245294609build26-1777585369814001SbV7","1768245294609build26-1777585425497001OifW","1768245294609build26-1777482299920001tTnQ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.386Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Qualcomm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776888767709001tN4X_20260512031421430.avif","ogTitle":"Run AI coding agents on-device with NPUs","ogDescription":"AI-powered coding agents are transforming developer workflows, but most run in the cloud, adding latency, cost, and privacy concerns. What if your coding agent ran entirely on your laptop’s NPU? In this session, you'll learn how Visual Studio AI Toolkit and Qualcomm’s Hexagon NPU enable fully on device agentic coding. Using optimized SLMs, you'll experience real time code generation with no cloud calls, delivering high performance per watt, longer battery life, and responsive developer UX.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner","SpeakerId":"1768245294609build26-1777482299920001tTnQ","RegistrantKey":"1768245294609build26-1777482299920001tTnQ"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777585369814001SbV7","RegistrantKey":"1768245294609build26-1777585369814001SbV7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777585425497001OifW","RegistrantKey":"1768245294609build26-1777585425497001OifW"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"2bc47037-7a6e-4a15-bc2a-b63702940d64","startDateTime":"2026-06-03T17:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"David Okeyode","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776295183936001WxNh","langLocale":"en-US","sessionId":"1776295183936001WxNh","sessionInstanceId":"1776295183936001WxNh","sessionCode":"LTG467","sessionCodeNormalized":"EN-US-LTG467","title":"Secure AI Agents in Azure: AI Gateway, Tools, and Trust","sortTitle":"secure ai agents in azure: ai gateway, tools, and trust","description":"See how to secure AI agents in Azure with controls developers can use immediately. In this live demo, we’ll protect agent-to-tool and agent-to-API interactions using Azure API Management as an AI gateway, govern MCP tools and registries with API Center, enforce identity and secret hygiene with Entra ID and Key Vault, and add observability for tracing and audit. You’ll learn practical ways to reduce prompt injection, over-permissioned tools, data leakage, and unsafe agent behavior.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:30 - 17:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776295183936001WxNh_20260512030356168.avif","speakerIds":["1768245294609build26-1769864099959001hxn4"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.892Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776295183936001WxNh_20260512030356168.avif","ogTitle":"Secure AI Agents in Azure: AI Gateway, Tools, and Trust","ogDescription":"See how to secure AI agents in Azure with controls developers can use immediately. In this live demo, we’ll protect agent-to-tool and agent-to-API interactions using Azure API Management as an AI gateway, govern MCP tools and registries with API Center, enforce identity and secret hygiene with Entra ID and Key Vault, and add observability for tracing and audit. You’ll learn practical ways to reduce prompt injection, over-permissioned tools, data leakage, and unsafe agent behavior.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1769864099959001hxn4","RegistrantKey":"1768245294609build26-1769864099959001hxn4"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"02b4c233-8026-47a4-8c84-8b4c2e21619c","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jason Berberich, Krishna Sunkavalli","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707747001rsJI","langLocale":"en-US","sessionId":"1774018707747001rsJI","sessionInstanceId":"1774018707747001rsJI","sessionCode":"LTG412","sessionCodeNormalized":"EN-US-LTG412","title":"Securing AI in production with Microsoft Defender for Cloud","sortTitle":"securing ai in production with microsoft defender for cloud","description":"As AI workloads move into production, security blind spots and cloud posture become critical risks. Learn how Microsoft Defender for Cloud secures AI workloads end‑to‑end—from infrastructure and data to AI services and agent‑based applications. Using a real‑world multi‑agent conversational platform as an example, the session shows how to identify risks, detect threats, and apply security best practices that enable teams to run AI systems in production with confidence and trust.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:10 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707747001rsJI_20260512021847930.avif","speakerIds":["1768245294609build26-1739217805395001TeRF","1768245294609build26-1751987253355001EQg3"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707747001rsJI_20260512021847930.avif","ogTitle":"Securing AI in production with Microsoft Defender for Cloud","ogDescription":"As AI workloads move into production, security blind spots and cloud posture become critical risks. Learn how Microsoft Defender for Cloud secures AI workloads end‑to‑end—from infrastructure and data to AI services and agent‑based applications. Using a real‑world multi‑agent conversational platform as an example, the session shows how to identify risks, detect threats, and apply security best practices that enable teams to run AI systems in production with confidence and trust.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751987253355001EQg3","RegistrantKey":"1768245294609build26-1751987253355001EQg3"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1739217805395001TeRF","RegistrantKey":"1768245294609build26-1739217805395001TeRF"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"GitHub Advanced Security","logicalValue":"GitHub Advanced Security"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"OFT","logicalValue":"OFT"}]},{"@search.score":1.0,"documentId":"44d4bc89-20c3-42e3-a8f9-551595632b0a","startDateTime":"2026-06-02T23:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"William Liang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018715938001rF6X","langLocale":"en-US","sessionId":"1774018715938001rF6X","sessionInstanceId":"1774018715938001rF6X","sessionCode":"LTG418","sessionCodeNormalized":"EN-US-LTG418","title":"Smarter training patterns for better AI models","sortTitle":"smarter training patterns for better ai models","description":"One-size-fits-all models rarely deliver top results. Discover how Foundry helps you get more from your data by offering advanced training patterns. Explore fine-tuning, reinforcement learning from traces, and powerful evaluation tools. Learn how observability, governance, and built-in productivity features speed up customization and cost reduction.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"23:30 - 23:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715938001rF6X_20260512022703687.avif","speakerIds":["1768245294609build26-1774284998627001p6H4"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.854Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018715938001rF6X_20260512022703687.avif","ogTitle":"Smarter training patterns for better AI models","ogDescription":"One-size-fits-all models rarely deliver top results. Discover how Foundry helps you get more from your data by offering advanced training patterns. Explore fine-tuning, reinforcement learning from traces, and powerful evaluation tools. Learn how observability, governance, and built-in productivity features speed up customization and cost reduction.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998627001p6H4","RegistrantKey":"1768245294609build26-1774284998627001p6H4"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"746318ef-0a28-471b-9c30-0f42049c1c66","startDateTime":"2026-06-03T17:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rohan Ganesh, Troy McBride, Vivaan Sawant, Advika Vuppala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774051017514001eEBs","langLocale":"en-US","sessionId":"1774051017514001eEBs","sessionInstanceId":"1774051017514001eEBs","sessionCode":"LTG443","sessionCodeNormalized":"EN-US-LTG443","title":"SpoilSafe: Imagine Cup Finalist Reducing Food Waste","sortTitle":"spoilsafe: imagine cup finalist reducing food waste","description":"SpoilSafe is an Azure-powered cold-chain intelligence platform that predicts food freshness in real time. Sensor data streams through Azure IoT Hub and is processed using Azure Data Explorer and Azure Synapse Analytics, while spoilage prediction models run on Azure AI Foundry. AI-driven insights help reduce food waste, improve safety, and enable smarter supply-chain decisions at scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"17:10 - 17:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051017514001eEBs_20260512023325128.avif","speakerIds":["1768245294609build26-1775792452897001CnyX","1768245294609build26-1775882816983001k1U7","1768245294609build26-1776040528117001fUtV","1768245294609build26-1776049446660001RE11"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.634Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774051017514001eEBs_20260512023325128.avif","ogTitle":"SpoilSafe: Imagine Cup Finalist Reducing Food Waste","ogDescription":"SpoilSafe is an Azure-powered cold-chain intelligence platform that predicts food freshness in real time. Sensor data streams through Azure IoT Hub and is processed using Azure Data Explorer and Azure Synapse Analytics, while spoilage prediction models run on Azure AI Foundry. AI-driven insights help reduce food waste, improve safety, and enable smarter supply-chain decisions at scale.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"C","logicalValue":"C"},{"displayValue":"TypeScript","logicalValue":"TypeScript"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776049446660001RE11","RegistrantKey":"1768245294609build26-1776049446660001RE11"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775882816983001k1U7","RegistrantKey":"1768245294609build26-1775882816983001k1U7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775792452897001CnyX","RegistrantKey":"1768245294609build26-1775792452897001CnyX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776040528117001fUtV","RegistrantKey":"1768245294609build26-1776040528117001fUtV"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure ML","logicalValue":"Azure ML"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"App Integration","logicalValue":"App Integration"}]},{"@search.score":1.0,"documentId":"6439adf0-111e-4242-9f4b-db6ace770458","startDateTime":"2026-06-02T20:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Damien DeHart","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776442760322001Fw9z","langLocale":"en-US","sessionId":"1776442760322001Fw9z","sessionInstanceId":"1776442760322001Fw9z","sessionCode":"LTGSP498","sessionCodeNormalized":"EN-US-LTGSP498","title":"Stop vulnerabilities in AI-generated code before they ship","sortTitle":"stop vulnerabilities in ai-generated code before they ship","description":"Apps ship in hours with Copilot and Cursor. Vibes are great, until the vulns show up. AI code introduces new risks: hallucinated packages and insecure patterns. This session breaks down Aikido's real-world research on why traditional AppSec fails 'vibe coding' and how attackers exploit this shift. We’ll cover AI vuln patterns, execution vectors, supply chain risks, and practical guardrails. Vibe coding is inevitable; shipping vulns isn’t. Walk away with a security blueprint.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:00 - 20:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T20:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776442760322001Fw9z_20260513131419335.avif","speakerIds":["1768245294609build26-1777553487548001KHz5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.382Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Aikido Security"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776442760322001Fw9z_20260513131419335.avif","ogTitle":"Stop vulnerabilities in AI-generated code before they ship","ogDescription":"Apps ship in hours with Copilot and Cursor. Vibes are great, until the vulns show up. AI code introduces new risks: hallucinated packages and insecure patterns. This session breaks down Aikido's real-world research on why traditional AppSec fails 'vibe coding' and how attackers exploit this shift. We’ll cover AI vuln patterns, execution vectors, supply chain risks, and practical guardrails. Vibe coding is inevitable; shipping vulns isn’t. Walk away with a security blueprint.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777553487548001KHz5","RegistrantKey":"1768245294609build26-1777553487548001KHz5"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/sess/1776442760322001Fw9z/supmat/Stop%20Vulnerabilities%20in%20AI-Generated%20Code%20-%20Aikido%20Security%20MSFT%20Build%20June%202026_1780426289032001VYCP.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Azure Copilot","logicalValue":"Azure Copilot"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"Platform Security","logicalValue":"Platform Security"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"6b3eb6da-82bf-4766-9492-d35a0d24b9cd","startDateTime":"2026-06-03T20:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roelant Dieben","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776293617419001kS86","langLocale":"en-US","sessionId":"1776293617419001kS86","sessionInstanceId":"1776293617419001kS86","sessionCode":"LTG464","sessionCodeNormalized":"EN-US-LTG464","title":"Stop Writing Prompts. Start Writing Specs.","sortTitle":"stop writing prompts. start writing specs.","description":"Prompt engineering got us started, but it often leads to fragile systems that are hard to maintain and do not inspire confidence. In this session, we explore a different path. Using GitHub Spec Kit we build an agent from a specification, treating the spec as the contract that defines capabilities, tools, and outputs. Instead of rewriting prompts, we evolve the system by evolving the spec.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:20 - 20:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293617419001kS86_20260512030315540.avif","speakerIds":["1768245294609build26-1770888477835001mAZA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.891Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776293617419001kS86_20260512030315540.avif","ogTitle":"Stop Writing Prompts. Start Writing Specs.","ogDescription":"Prompt engineering got us started, but it often leads to fragile systems that are hard to maintain and do not inspire confidence. In this session, we explore a different path. Using GitHub Spec Kit we build an agent from a specification, treating the spec as the contract that defines capabilities, tools, and outputs. Instead of rewriting prompts, we evolve the system by evolving the spec.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770888477835001mAZA","RegistrantKey":"1768245294609build26-1770888477835001mAZA"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"GitHub Enterprise","logicalValue":"GitHub Enterprise"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"MVP","logicalValue":"MVP"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"cda409b1-61d5-4155-9c9e-93ed77b4bdf6","startDateTime":"2026-06-03T18:50:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Will Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774393260065001kyWv","langLocale":"en-US","sessionId":"1774393260065001kyWv","sessionInstanceId":"1774393260065001kyWv","sessionCode":"LTGSP487","sessionCodeNormalized":"EN-US-LTGSP487","title":"Streamline your end-to-end developer workflow with AI","sortTitle":"streamline your end-to-end developer workflow with ai","description":"Most developers want to ship their own ideas, but the friction between coding, communication, and release slows them down. In this session, follow a real daily workflow building Clova—from ideation to launch—using GitHub Copilot in VS Code, Microsoft Teams, Logitech MX tools, and AI-assisted workflows. Learn practical patterns to streamline work, reduce friction, and accelerate execution.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"18:50 - 19:05","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:05:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774393260065001kyWv_20260520160444633.avif","speakerIds":["1768245294609build26-1776428358928001Ukpm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.639Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Logitech"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774393260065001kyWv_20260520160444633.avif","ogTitle":"Streamline your end-to-end developer workflow with AI","ogDescription":"Most developers want to ship their own ideas, but the friction between coding, communication, and release slows them down. In this session, follow a real daily workflow building Clova—from ideation to launch—using GitHub Copilot in VS Code, Microsoft Teams, Logitech MX tools, and AI-assisted workflows. Learn practical patterns to streamline work, reduce friction, and accelerate execution.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776428358928001Ukpm","RegistrantKey":"1768245294609build26-1776428358928001Ukpm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"a3ea88e9-9663-42de-b016-4a7c3d2600ed","startDateTime":"2026-06-02T19:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Paul Yu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018707087001rg4W","langLocale":"en-US","sessionId":"1774018707087001rg4W","sessionInstanceId":"1774018707087001rg4W","sessionCode":"LTG408","sessionCodeNormalized":"EN-US-LTG408","title":"The developer experience with AKS Desktop and VS Code","sortTitle":"the developer experience with aks desktop and vs code","description":"Getting an app from local dev or on‑prem to AKS can be hard to make secure, repeatable, and easy to run. This session shows how AKS Desktop and VS Code speed the developer‑to‑AKS journey, from local Kubernetes workflows and containerisation to CI/CD and safe deployments, using built‑in Kubernetes best practices for reliable operations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:20 - 19:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T19:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707087001rg4W_20260512021813557.avif","speakerIds":["1768245294609build26-1774284994279001pbAB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018707087001rg4W_20260512021813557.avif","ogTitle":"The developer experience with AKS Desktop and VS Code","ogDescription":"Getting an app from local dev or on‑prem to AKS can be hard to make secure, repeatable, and easy to run. This session shows how AKS Desktop and VS Code speed the developer‑to‑AKS journey, from local Kubernetes workflows and containerisation to CI/CD and safe deployments, using built‑in Kubernetes best practices for reliable operations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[{"displayValue":"C","logicalValue":"C"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284994279001pbAB","RegistrantKey":"1768245294609build26-1774284994279001pbAB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"}]},{"@search.score":1.0,"documentId":"59dda0da-6ee7-448b-b68c-c951479e29a5","startDateTime":"2026-06-03T21:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jason Fisher, Nazmus Sakib","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018720565001rjwn","langLocale":"en-US","sessionId":"1774018720565001rjwn","sessionInstanceId":"1774018720565001rjwn","sessionCode":"LTG436","sessionCodeNormalized":"EN-US-LTG436","title":"The Windows Security Features That Matter Most for Developers","sortTitle":"the windows security features that matter most for developers","description":"Windows security foundations are evolving at the OS level to support post‑quantum resilience and emerging agentic workflows.  We’ll cover what developers can start adopting now to help raise the security bar across the ecosystem while preserving application compatibility and minimizing friction.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:20 - 21:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720565001rjwn_20260512023124373.avif","speakerIds":["1768245294609build26-1775493934666001qrTz","1768245294609build26-1774285003098001pwv1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.633Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720565001rjwn_20260512023124373.avif","ogTitle":"The Windows Security Features That Matter Most for Developers","ogDescription":"Windows security foundations are evolving at the OS level to support post‑quantum resilience and emerging agentic workflows.  We’ll cover what developers can start adopting now to help raise the security bar across the ecosystem while preserving application compatibility and minimizing friction.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285003098001pwv1","RegistrantKey":"1768245294609build26-1774285003098001pwv1"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775493934666001qrTz","RegistrantKey":"1768245294609build26-1775493934666001qrTz"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows Security","logicalValue":"Windows Security"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"AppTrust","logicalValue":"AppTrust"},{"displayValue":"Platform Security","logicalValue":"Platform Security"},{"displayValue":"Post Quantum Cryptography","logicalValue":"Post Quantum Cryptography"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"}]},{"@search.score":1.0,"documentId":"c44c5cdc-d822-452d-813d-0977e2dd3718","startDateTime":"2026-06-03T21:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Taylor Lykins","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774392351303001wESh","langLocale":"en-US","sessionId":"1774392351303001wESh","sessionInstanceId":"1774392351303001wESh","sessionCode":"LTGSP483","sessionCodeNormalized":"EN-US-LTGSP483","title":"Turn documentation into a context layer for production AI","sortTitle":"turn documentation into a context layer for production ai","description":"This session examines a paradigm shift in AI transformation readiness: documentation as infrastructure. It reimagines an organization’s operating system, where processes, architecture standards, and workflows move from static archives to version‑controlled, computable logic. Learn how teams build a context layer that feeds AI systems with intent, standards, and constraints to improve reliability, scale, and real‑world outcomes.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:00 - 21:15","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:15:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392351303001wESh_20260518110631194.avif","speakerIds":["1768245294609build26-1776800276473001XqU7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.637Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Lucid Software"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774392351303001wESh_20260518110631194.avif","ogTitle":"Turn documentation into a context layer for production AI","ogDescription":"This session examines a paradigm shift in AI transformation readiness: documentation as infrastructure. It reimagines an organization’s operating system, where processes, architecture standards, and workflows move from static archives to version‑controlled, computable logic. Learn how teams build a context layer that feeds AI systems with intent, standards, and constraints to improve reliability, scale, and real‑world outcomes.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776800276473001XqU7","RegistrantKey":"1768245294609build26-1776800276473001XqU7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Software Development Company","logicalValue":"Software Development Company"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"56853c97-e682-4bbc-b2c9-39e6305f8f50","startDateTime":"2026-06-03T20:40:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gulsimo Osimi, Saumil Shrivastava","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018716247001rclm","langLocale":"en-US","sessionId":"1774018716247001rclm","sessionInstanceId":"1774018716247001rclm","sessionCode":"LTG419","sessionCodeNormalized":"EN-US-LTG419","title":"Turn ideas into AI applications with Microsoft Foundry Labs","sortTitle":"turn ideas into ai applications with microsoft foundry labs","description":"Developers need a fast path from ideas to real AI applications. In this lightning talk, Microsoft introduces Foundry Labs 2.0, a new experience for rapid AI prototyping. Through a live demo, developers will explore domain specific models and data, experiment in interactive playgrounds, and compose models, data, and agents into a working application. The session highlights an end-to-end workflow that accelerates experimentation into deployable AI solutions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"20:40 - 20:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T20:55:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018716247001rclm_20260512022712762.avif","speakerIds":["1768245294609build26-1777308405664001CTUG","1768245294609build26-1777308294271001nvnK"],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.855Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018716247001rclm_20260512022712762.avif","ogTitle":"Turn ideas into AI applications with Microsoft Foundry Labs","ogDescription":"Developers need a fast path from ideas to real AI applications. In this lightning talk, Microsoft introduces Foundry Labs 2.0, a new experience for rapid AI prototyping. Through a live demo, developers will explore domain specific models and data, experiment in interactive playgrounds, and compose models, data, and agents into a working application. The session highlights an end-to-end workflow that accelerates experimentation into deployable AI solutions.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777308294271001nvnK","RegistrantKey":"1768245294609build26-1777308294271001nvnK"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777308405664001CTUG","RegistrantKey":"1768245294609build26-1777308405664001CTUG"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/foundrylabs","backpackLink":null,"label":"Learn more about Foundry Labs","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"}]},{"@search.score":1.0,"documentId":"fd13cd87-6beb-4b54-9c77-91a3163e7f69","startDateTime":"2026-06-03T19:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Alex Pshul","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776290734305001x7s6","langLocale":"en-US","sessionId":"1776290734305001x7s6","sessionInstanceId":"1776290734305001x7s6","sessionCode":"LTG446","sessionCodeNormalized":"EN-US-LTG446","title":"TypeScript LangChain agents on Microsoft Foundry Agents Service","sortTitle":"typescript langchain agents on microsoft foundry agents service","description":"Build TypeScript LangChain agents on Microsoft Foundry Agent Service without giving up your existing stack. You get managed agent runtime, tools, tracing, and enterprise-ready deployment while staying in TypeScript. Now it’s possible to connect LangChain/LangGraph to Microsoft Foundry by using the hosted agents offering and run real agent workflows, not just chat demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"19:10 - 19:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776290734305001x7s6_20260512025932076.avif","speakerIds":["1768245294609build26-1757087970341001aGw6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:54.887Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776290734305001x7s6_20260512025932076.avif","ogTitle":"TypeScript LangChain agents on Microsoft Foundry Agents Service","ogDescription":"Build TypeScript LangChain agents on Microsoft Foundry Agent Service without giving up your existing stack. You get managed agent runtime, tools, tracing, and enterprise-ready deployment while staying in TypeScript. Now it’s possible to connect LangChain/LangGraph to Microsoft Foundry by using the hosted agents offering and run real agent workflows, not just chat demos.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1757087970341001aGw6","RegistrantKey":"1768245294609build26-1757087970341001aGw6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"f695cba4-ac28-40c4-8515-5cd60ee76404","startDateTime":"2026-06-03T16:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nikola Metulev, Beth Pan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018720726001rkqn","langLocale":"en-US","sessionId":"1774018720726001rkqn","sessionInstanceId":"1774018720726001rkqn","sessionCode":"LTG437","sessionCodeNormalized":"EN-US-LTG437","title":"Use agents to build WinUI3 apps","sortTitle":"use agents to build winui3 apps","description":"Learn how agent-based development workflows can be used to build modern Windows applications with WinUI 3. In this demo, we’ll showcase new platform capabilities to enable agents to assist with application design, architecture, implementation, and testing. See how developers can leverage these capabilities to create new WinUI 3 apps, improve existing applications, or migrate legacy apps to a modern Windows UI stack using intelligent automation across the development lifecycle.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"16:30 - 16:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T16:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720726001rkqn_20260512023133243.avif","speakerIds":["1768245294609build26-1774886013101001R9mk","1768245294609build26-1775078130718001cIWp"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.858Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018720726001rkqn_20260512023133243.avif","ogTitle":"Use agents to build WinUI3 apps","ogDescription":"Learn how agent-based development workflows can be used to build modern Windows applications with WinUI 3. In this demo, we’ll showcase new platform capabilities to enable agents to assist with application design, architecture, implementation, and testing. See how developers can leverage these capabilities to create new WinUI 3 apps, improve existing applications, or migrate legacy apps to a modern Windows UI stack using intelligent automation across the development lifecycle.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774886013101001R9mk","RegistrantKey":"1768245294609build26-1774886013101001R9mk"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775078130718001cIWp","RegistrantKey":"1768245294609build26-1775078130718001cIWp"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"WinUI","logicalValue":"WinUI"},{"displayValue":"Skills","logicalValue":"Skills"}]},{"@search.score":1.0,"documentId":"8c31e2bf-8689-413a-9cf9-6d83b71ad365","startDateTime":"2026-06-03T00:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Justin Chen","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711240001rnkh","langLocale":"en-US","sessionId":"1774018711240001rnkh","sessionInstanceId":"1774018711240001rnkh","sessionCode":"LTG400","sessionCodeNormalized":"EN-US-LTG400","title":"VS Code features that would have saved you last week","sortTitle":"vs code features that would have saved you last week","description":"VS Code ships a lot of features and some of the best ones fly under the radar. This talk is a rapid-fire tour of the ones that don’t make the headlines but save real time: sticky scroll, timeline view, built-in port forwarding, terminal IntelliSense, the merge editor, and more.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:10 - 00:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711240001rnkh_20260512022326646.avif","speakerIds":["1768245294609build26-1774285001041001p103"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.789Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711240001rnkh_20260512022326646.avif","ogTitle":"VS Code features that would have saved you last week","ogDescription":"VS Code ships a lot of features and some of the best ones fly under the radar. This talk is a rapid-fire tour of the ones that don’t make the headlines but save real time: sticky scroll, timeline view, built-in port forwarding, terminal IntelliSense, the merge editor, and more.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285001041001p103","RegistrantKey":"1768245294609build26-1774285001041001p103"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"3e8930d5-5b97-40eb-95af-fd3c2b444a84","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vinay Perneti Perneti","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774544028291001LyOv","langLocale":"en-US","sessionId":"1774544028291001LyOv","sessionInstanceId":"1774544028291001LyOv","sessionCode":"LTGSP489","sessionCodeNormalized":"EN-US-LTGSP489","title":"We thought AI transformation was about adopting agents. We were wrong.","sortTitle":"we thought ai transformation was about adopting agents. we were wrong.","description":"We have data showing AI agents are improving at a super-exponential rate. If this continues, we'll likely have agents 10x more capable than the ones we had in November 2025 — within a year.\n\nIn this talk, we'll cover three myths about AI-Native transformations and how Augment transformed itself over the last three months. We've been blogging openly about this journey. You'll get the whole story and the learnings — so you can do the same for your organization.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"21:30 - 21:45","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:45:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774544028291001LyOv_20260603024816004.avif","speakerIds":["1768245294609build26-1778525837808001aUtC"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.842Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Augment Code"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774544028291001LyOv_20260603024816004.avif","ogTitle":"We thought AI transformation was about adopting agents. We were wrong.","ogDescription":"We have data showing AI agents are improving at a super-exponential rate. If this continues, we'll likely have agents 10x more capable than the ones we had in November 2025 — within a year.\n\nIn this talk, we'll cover three myths about AI-Native transformations and how Augment transformed itself over the last three months. We've been blogging openly about this journey. You'll get the whole story and the learnings — so you can do the same for your organization.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 1","logicalValue":"Gateway Pavilion, Level 2, Theater 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778525837808001aUtC","RegistrantKey":"1768245294609build26-1778525837808001aUtC"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"9703b5bb-1a5e-4235-87b5-e61d1f460b56","startDateTime":"2026-06-02T22:10:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maria Bledsoe","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018702855001rX4o","langLocale":"en-US","sessionId":"1774018702855001rX4o","sessionInstanceId":"1774018702855001rX4o","sessionCode":"LTG421","sessionCodeNormalized":"EN-US-LTG421","title":"What the shift from cloud apps to agentic systems means for developers","sortTitle":"what the shift from cloud apps to agentic systems means for developers","description":"AI is changing how applications and agents are built and operated. This lightning talk explains how agentic systems reason over context, take multi‑step actions, and adapt over time—and what that means for developers. We’ll cover how design, building, delivery, and operations evolve as apps become agent‑driven, and how Microsoft’s integrated agentic platform is able to support apps and agents responsibly in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:10 - 22:25","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:25:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018702855001rX4o_20260512021353149.avif","speakerIds":["1768245294609build26-1759345812677001kVEa"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.458Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018702855001rX4o_20260512021353149.avif","ogTitle":"What the shift from cloud apps to agentic systems means for developers","ogDescription":"AI is changing how applications and agents are built and operated. This lightning talk explains how agentic systems reason over context, take multi‑step actions, and adapt over time—and what that means for developers. We’ll cover how design, building, delivery, and operations evolve as apps become agent‑driven, and how Microsoft’s integrated agentic platform is able to support apps and agents responsibly in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 2","logicalValue":"Gateway Pavilion, Level 2, Theater 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1759345812677001kVEa","RegistrantKey":"1768245294609build26-1759345812677001kVEa"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"5c547462-1814-4a9b-9b75-ea5be59fb328","startDateTime":"2026-06-03T22:20:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Harald Kirschner, Courtney Webster","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774018711578001riff","langLocale":"en-US","sessionId":"1774018711578001riff","sessionInstanceId":"1774018711578001riff","sessionCode":"LTG402","sessionCodeNormalized":"EN-US-LTG402","title":"Why GitHub Copilot misses context (and how to fix it)","sortTitle":"why github copilot misses context (and how to fix it)","description":"Out of the box, GitHub Copilot doesn't know your conventions, architecture opinions, or how you ship. But you can teach it. Set up shared rules and reusable prompts so Copilot follows your standards. Build custom agents and skills for workflows your team actually runs. Dial in settings that scale from five engineers to five hundred. Rules, skills, prompt files, custom agents — see how they fit and when to reach for each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"22:20 - 22:35","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:35:00Z","durationInMinutes":15,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711578001riff_20260512022350205.avif","speakerIds":["1768245294609build26-1774285000231001pWGh","1768245294609build26-1774284995211001pJD3"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.79Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774018711578001riff_20260512022350205.avif","ogTitle":"Why GitHub Copilot misses context (and how to fix it)","ogDescription":"Out of the box, GitHub Copilot doesn't know your conventions, architecture opinions, or how you ship. But you can teach it. Set up shared rules and reusable prompts so Copilot follows your standards. Build custom agents and skills for workflows your team actually runs. Dial in settings that scale from five engineers to five hundred. Rules, skills, prompt files, custom agents — see how they fit and when to reach for each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater 3","logicalValue":"Gateway Pavilion, Level 2, Theater 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000231001pWGh","RegistrantKey":"1768245294609build26-1774285000231001pWGh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995211001pJD3","RegistrantKey":"1768245294609build26-1774284995211001pJD3"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"8a61210e-90ca-4016-9571-88e00847e7e3","startDateTime":"2026-06-04T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Harald Kirschner, Courtney Webster","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778263231219001ma0q","langLocale":"en-US","sessionId":"1778263231219001ma0q","sessionInstanceId":"1778263231219001ma0q","sessionCode":"LTG402-R1","sessionCodeNormalized":"EN-US-LTG402-R1","title":"Why GitHub Copilot misses context (and how to fix it) (Repeat)","sortTitle":"why github copilot misses context (and how to fix it) (repeat)","description":"Out of the box, GitHub Copilot doesn't know your conventions, architecture opinions, or how you ship. But you can teach it. Set up shared rules and reusable prompts so Copilot follows your standards. Build custom agents and skills for workflows your team actually runs. Dial in settings that scale from five engineers to five hundred. Rules, skills, prompt files, custom agents — see how they fit and when to reach for each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":6,"TimeSlot":"00:30 - 00:55","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:55:00Z","durationInMinutes":25,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263231219001ma0q_20260512031957247.avif","speakerIds":["1768245294609build26-1774285000231001pWGh","1768245294609build26-1774284995211001pJD3"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.775Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778263231219001ma0q_20260512031957247.avif","ogTitle":"Why GitHub Copilot misses context (and how to fix it) (Repeat)","ogDescription":"Out of the box, GitHub Copilot doesn't know your conventions, architecture opinions, or how you ship. But you can teach it. Set up shared rules and reusable prompts so Copilot follows your standards. Build custom agents and skills for workflows your team actually runs. Dial in settings that scale from five engineers to five hundred. Rules, skills, prompt files, custom agents — see how they fit and when to reach for each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Lightning Talk","logicalValue":"Lightning Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Theater C","logicalValue":"Gateway Pavilion, Level 2, Theater C"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285000231001pWGh","RegistrantKey":"1768245294609build26-1774285000231001pWGh"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284995211001pJD3","RegistrantKey":"1768245294609build26-1774284995211001pJD3"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"8bf5e718-ecca-4348-a29b-f8456e8d54a3","startDateTime":"2026-06-03T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Brian Benz","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493284181001kEfV","langLocale":"en-US","sessionId":"1774493284181001kEfV","sessionInstanceId":"1774493284181001kEfV","sessionCode":"TT666A","sessionCodeNormalized":"EN-US-TT666A","title":"Adding AI to Real Apps: The Agentic Way","sortTitle":"adding ai to real apps: the agentic way","description":"Most teams aren’t building AI apps from scratch—they’re adding AI to applications that already exist. And increasingly, they’re not writing that integration themselves. This Table Talk explores how agentic development workflows—Copilot CLI, agent mode, and cloud coding agents—are reshaping how AI capabilities get added to real apps. We’ll discuss practical integration patterns, where agents accelerate the work, and where they still need a human in the loop. Bring your own modernization stories as we compare what’s working, what isn’t, and where to start.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284181001kEfV_20260512153225385.avif","speakerIds":["1768245294609build26-1776199101971001JvYe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.839Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284181001kEfV_20260512153225385.avif","ogTitle":"Adding AI to Real Apps: The Agentic Way","ogDescription":"Most teams aren’t building AI apps from scratch—they’re adding AI to applications that already exist. And increasingly, they’re not writing that integration themselves. This Table Talk explores how agentic development workflows—Copilot CLI, agent mode, and cloud coding agents—are reshaping how AI capabilities get added to real apps. We’ll discuss practical integration patterns, where agents accelerate the work, and where they still need a human in the loop. Bring your own modernization stories as we compare what’s working, what isn’t, and where to start.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776199101971001JvYe","RegistrantKey":"1768245294609build26-1776199101971001JvYe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"3c7b7bd4-6290-4d9c-bbec-7fe051e583aa","startDateTime":"2026-06-03T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayan Gupta","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778525089414001n9ip","langLocale":"en-US","sessionId":"1778525089414001n9ip","sessionInstanceId":"1778525089414001n9ip","sessionCode":"TT666B","sessionCodeNormalized":"EN-US-TT666B","title":"Adding AI to Real Apps: The Agentic Way","sortTitle":"adding ai to real apps: the agentic way","description":"Most teams aren’t building AI apps from scratch—they’re adding AI to applications that already exist. And increasingly, they’re not writing that integration themselves. This Table Talk explores how agentic development workflows—Copilot CLI, agent mode, and cloud coding agents—are reshaping how AI capabilities get added to real apps. We’ll discuss practical integration patterns, where agents accelerate the work, and where they still need a human in the loop. Bring your own modernization stories as we compare what’s working, what isn’t, and where to start.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778525089414001n9ip_20260512152903321.avif","speakerIds":["1768245294609build26-1767923733240001J5fY"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.776Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778525089414001n9ip_20260512152903321.avif","ogTitle":"Adding AI to Real Apps: The Agentic Way","ogDescription":"Most teams aren’t building AI apps from scratch—they’re adding AI to applications that already exist. And increasingly, they’re not writing that integration themselves. This Table Talk explores how agentic development workflows—Copilot CLI, agent mode, and cloud coding agents—are reshaping how AI capabilities get added to real apps. We’ll discuss practical integration patterns, where agents accelerate the work, and where they still need a human in the loop. Bring your own modernization stories as we compare what’s working, what isn’t, and where to start.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767923733240001J5fY","RegistrantKey":"1768245294609build26-1767923733240001J5fY"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"c26dc843-511d-468c-84c4-eae44ad22aa3","startDateTime":"2026-06-03T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Carlo","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493285851001kTlq","langLocale":"en-US","sessionId":"1774493285851001kTlq","sessionInstanceId":"1774493285851001kTlq","sessionCode":"TT677","sessionCodeNormalized":"EN-US-TT677","title":"Agentic build experiences with Microsoft Fabric","sortTitle":"agentic build experiences with microsoft fabric","description":"Explore how GitHub Copilot and MCP servers can accelerate your modeling, reporting, and Microsoft Fabric deployment workflows. Unpack new agentic developer experiences, see powerful prompts to enhance Power BI and Fabric, and receive practical code snippets, and proven techniques to boost your efficiency right away. Learn agentic systems combine AI with practical know-how skills.md AI dev Come ready to chat and discover why this is the next wave in Fabric development!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"16:15 - 17:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285851001kTlq_20260512024323275.avif","speakerIds":["1768245294609build26-1776293231988001WPJb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.841Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285851001kTlq_20260512024323275.avif","ogTitle":"Agentic build experiences with Microsoft Fabric","ogDescription":"Explore how GitHub Copilot and MCP servers can accelerate your modeling, reporting, and Microsoft Fabric deployment workflows. Unpack new agentic developer experiences, see powerful prompts to enhance Power BI and Fabric, and receive practical code snippets, and proven techniques to boost your efficiency right away. Learn agentic systems combine AI with practical know-how skills.md AI dev Come ready to chat and discover why this is the next wave in Fabric development!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293231988001WPJb","RegistrantKey":"1768245294609build26-1776293231988001WPJb"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/TT677%20-%20Agentic%20Build%20Experiences%20with%20Microsoft%20Fabric_1779990205743001Rbce%20%281%29_1780511473412001bEn2.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"MVP","logicalValue":"MVP"}]},{"@search.score":1.0,"documentId":"ed1ab040-de41-40cd-ac05-f5ef27891f8c","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Carlo","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779813319604001Sbly","langLocale":"en-US","sessionId":"1779813319604001Sbly","sessionInstanceId":"1779813319604001Sbly","sessionCode":"TT677-R1","sessionCodeNormalized":"EN-US-TT677-R1","title":"Agentic build experiences with Microsoft Fabric","sortTitle":"agentic build experiences with microsoft fabric","description":"Explore how GitHub Copilot and MCP servers can accelerate your modeling, reporting, and Microsoft Fabric deployment workflows. Unpack new agentic developer experiences, see powerful prompts to enhance Power BI and Fabric, and receive practical code snippets, and proven techniques to boost your efficiency right away. Learn agentic systems combine AI with practical know-how skills.md AI dev Come ready to chat and discover why this is the next wave in Fabric development!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779813319604001Sbly_20260527112712599.avif","speakerIds":["1768245294609build26-1776293231988001WPJb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.238Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779813319604001Sbly_20260527112712599.avif","ogTitle":"Agentic build experiences with Microsoft Fabric","ogDescription":"Explore how GitHub Copilot and MCP servers can accelerate your modeling, reporting, and Microsoft Fabric deployment workflows. Unpack new agentic developer experiences, see powerful prompts to enhance Power BI and Fabric, and receive practical code snippets, and proven techniques to boost your efficiency right away. Learn agentic systems combine AI with practical know-how skills.md AI dev Come ready to chat and discover why this is the next wave in Fabric development!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293231988001WPJb","RegistrantKey":"1768245294609build26-1776293231988001WPJb"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/static/staticfile/staticfile/TT677%20-%20Agentic%20Build%20Experiences%20with%20Microsoft%20Fabric_1779990205743001Rbce%20%281%29_1780511473412001bEn2.pptx","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Developers","logicalValue":"App Developers"}]},{"@search.score":1.0,"documentId":"e8fd0a9d-c273-4cf4-aa70-042cbf81f1cc","startDateTime":"2026-06-02T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ofek Ben Eliezer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493286011001k8PS","langLocale":"en-US","sessionId":"1774493286011001k8PS","sessionInstanceId":"1774493286011001k8PS","sessionCode":"TT678","sessionCodeNormalized":"EN-US-TT678","title":"AI Observability: What You Can’t See (and Why It Breaks)","sortTitle":"ai observability: what you can’t see (and why it breaks)","description":"Traditional monitoring can’t catch LLM hallucinations or hidden token costs. In this discussion, we’ll move beyond \"green\" status lights to the architectural reality of running AI at scale. We will dive into Microsoft Foundry to explore correlating signals across infra, data, and model layers. Bring your production \"war stories\" as we debate how to use Azure AI Evaluation and tracing to build systems that are truly debuggable, measurable, and continuously improvable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286011001k8PS_20260512024335648.avif","speakerIds":["1768245294609build26-1776296228053001Ld8t"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.841Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286011001k8PS_20260512024335648.avif","ogTitle":"AI Observability: What You Can’t See (and Why It Breaks)","ogDescription":"Traditional monitoring can’t catch LLM hallucinations or hidden token costs. In this discussion, we’ll move beyond \"green\" status lights to the architectural reality of running AI at scale. We will dive into Microsoft Foundry to explore correlating signals across infra, data, and model layers. Bring your production \"war stories\" as we debate how to use Azure AI Evaluation and tracing to build systems that are truly debuggable, measurable, and continuously improvable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296228053001Ld8t","RegistrantKey":"1768245294609build26-1776296228053001Ld8t"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"af632643-8787-44cb-a1c4-aa850a9d8ce6","startDateTime":"2026-06-03T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ofek Ben Eliezer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779812645028001qU6Y","langLocale":"en-US","sessionId":"1779812645028001qU6Y","sessionInstanceId":"1779812645028001qU6Y","sessionCode":"TT678-R1","sessionCodeNormalized":"EN-US-TT678-R1","title":"AI Observability: What You Can’t See (and Why It Breaks)","sortTitle":"ai observability: what you can’t see (and why it breaks)","description":"Traditional monitoring can’t catch LLM hallucinations or hidden token costs. In this discussion, we’ll move beyond \"green\" status lights to the architectural reality of running AI at scale. We will dive into Microsoft Foundry to explore correlating signals across infra, data, and model layers. Bring your production \"war stories\" as we debate how to use Azure AI Evaluation and tracing to build systems that are truly debuggable, measurable, and continuously improvable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812645028001qU6Y_20260527112621252.avif","speakerIds":["1768245294609build26-1776296228053001Ld8t"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.238Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812645028001qU6Y_20260527112621252.avif","ogTitle":"AI Observability: What You Can’t See (and Why It Breaks)","ogDescription":"Traditional monitoring can’t catch LLM hallucinations or hidden token costs. In this discussion, we’ll move beyond \"green\" status lights to the architectural reality of running AI at scale. We will dive into Microsoft Foundry to explore correlating signals across infra, data, and model layers. Bring your production \"war stories\" as we debate how to use Azure AI Evaluation and tracing to build systems that are truly debuggable, measurable, and continuously improvable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296228053001Ld8t","RegistrantKey":"1768245294609build26-1776296228053001Ld8t"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"0bbc1ead-48f5-4f93-b8fc-8307d239e49a","startDateTime":"2026-06-03T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"James Ray","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493198742001nbqP","langLocale":"en-US","sessionId":"1774493198742001nbqP","sessionInstanceId":"1774493198742001nbqP","sessionCode":"TT641","sessionCodeNormalized":"EN-US-TT641","title":"AI safety at scale with the MAI team","sortTitle":"ai safety at scale with the mai team","description":"Shipping AI safely is an ongoing process of experimentation, investigation, creativity, and design. We’ll share how we've built safety systems for large-scale nondeterministic systems – making them robust to emerging harms and adversarial actors, and how we tackle trade-offs like over-refusal. Expect a discussion-driven session on building AI that’s flexible, reliable, and accountable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198742001nbqP_20260512153137232.avif","speakerIds":["1768245294609build26-1777903976787001e7jT"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.793Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198742001nbqP_20260512153137232.avif","ogTitle":"AI safety at scale with the MAI team","ogDescription":"Shipping AI safely is an ongoing process of experimentation, investigation, creativity, and design. We’ll share how we've built safety systems for large-scale nondeterministic systems – making them robust to emerging harms and adversarial actors, and how we tackle trade-offs like over-refusal. Expect a discussion-driven session on building AI that’s flexible, reliable, and accountable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777903976787001e7jT","RegistrantKey":"1768245294609build26-1777903976787001e7jT"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"ea53ea69-b3a3-409f-ba12-478b46b71006","startDateTime":"2026-06-03T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"James Ray","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778003701552001jZio","langLocale":"en-US","sessionId":"1778003701552001jZio","sessionInstanceId":"1778003701552001jZio","sessionCode":"TT641-R1","sessionCodeNormalized":"EN-US-TT641-R1","title":"AI safety at scale with the MAI team","sortTitle":"ai safety at scale with the mai team","description":"Shipping AI safely is an ongoing process of experimentation, investigation, creativity, and design. We’ll share how we've built safety systems for large-scale nondeterministic systems – making them robust to emerging harms and adversarial actors, and how we tackle trade-offs like over-refusal. Expect a discussion-driven session on building AI that’s flexible, reliable, and accountable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778003701552001jZio_20260512153325772.avif","speakerIds":["1768245294609build26-1777903976787001e7jT"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.768Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778003701552001jZio_20260512153325772.avif","ogTitle":"AI safety at scale with the MAI team","ogDescription":"Shipping AI safely is an ongoing process of experimentation, investigation, creativity, and design. We’ll share how we've built safety systems for large-scale nondeterministic systems – making them robust to emerging harms and adversarial actors, and how we tackle trade-offs like over-refusal. Expect a discussion-driven session on building AI that’s flexible, reliable, and accountable.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777903976787001e7jT","RegistrantKey":"1768245294609build26-1777903976787001e7jT"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"d03d6ed0-22ab-4418-af64-e8aca910eb10","startDateTime":"2026-06-02T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nagkumar Arkalgud, Mikhail Simin, Hanchi Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493198590001n4Qd","langLocale":"en-US","sessionId":"1774493198590001n4Qd","sessionInstanceId":"1774493198590001n4Qd","sessionCode":"TT640","sessionCodeNormalized":"EN-US-TT640","title":"Any agent, any cloud: Observability patterns that scale","sortTitle":"any agent, any cloud: observability patterns that scale","description":"Agents ship across frameworks and clouds—but telemetry is often fragmented. Let’s map an observability plan: what to trace (model + tool calls), what to measure (latency, cost, quality), and how to optimize (using trace-linked insights). Bring your scenario; we’ll compare a few viable patterns, debate tradeoffs, and leave with a practical checklist of next actions for your stack.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198590001n4Qd_20260512024048634.avif","speakerIds":["1768245294609build26-1776181207688001qvOn","1768245294609build26-1776181296090001Jn71","1768245294609build26-1774284998317001piuN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.808Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198590001n4Qd_20260512024048634.avif","ogTitle":"Any agent, any cloud: Observability patterns that scale","ogDescription":"Agents ship across frameworks and clouds—but telemetry is often fragmented. Let’s map an observability plan: what to trace (model + tool calls), what to measure (latency, cost, quality), and how to optimize (using trace-linked insights). Bring your scenario; we’ll compare a few viable patterns, debate tradeoffs, and leave with a practical checklist of next actions for your stack.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998317001piuN","RegistrantKey":"1768245294609build26-1774284998317001piuN"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181207688001qvOn","RegistrantKey":"1768245294609build26-1776181207688001qvOn"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181296090001Jn71","RegistrantKey":"1768245294609build26-1776181296090001Jn71"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Tracing","logicalValue":"Tracing"}]},{"@search.score":1.0,"documentId":"82836dda-e79b-4368-97ea-71800fdc0e35","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nagkumar Arkalgud, Mikhail Simin, Hanchi Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777930100455001LxFj","langLocale":"en-US","sessionId":"1777930100455001LxFj","sessionInstanceId":"1777930100455001LxFj","sessionCode":"TT640-R1","sessionCodeNormalized":"EN-US-TT640-R1","title":"Any agent, any cloud: Observability patterns that scale","sortTitle":"any agent, any cloud: observability patterns that scale","description":"Agents ship across frameworks and clouds—but telemetry is often fragmented. Let’s map an observability plan: what to trace (model + tool calls), what to measure (latency, cost, quality), and how to optimize (using trace-linked insights). Bring your scenario; we’ll compare a few viable patterns, debate tradeoffs, and leave with a practical checklist of next actions for your stack.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777930100455001LxFj_20260512031651763.avif","speakerIds":["1768245294609build26-1776181207688001qvOn","1768245294609build26-1776181296090001Jn71","1768245294609build26-1774284998317001piuN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.39Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777930100455001LxFj_20260512031651763.avif","ogTitle":"Any agent, any cloud: Observability patterns that scale","ogDescription":"Agents ship across frameworks and clouds—but telemetry is often fragmented. Let’s map an observability plan: what to trace (model + tool calls), what to measure (latency, cost, quality), and how to optimize (using trace-linked insights). Bring your scenario; we’ll compare a few viable patterns, debate tradeoffs, and leave with a practical checklist of next actions for your stack.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284998317001piuN","RegistrantKey":"1768245294609build26-1774284998317001piuN"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181207688001qvOn","RegistrantKey":"1768245294609build26-1776181207688001qvOn"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776181296090001Jn71","RegistrantKey":"1768245294609build26-1776181296090001Jn71"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Tracing","logicalValue":"Tracing"}]},{"@search.score":1.0,"documentId":"28df4b16-1b05-4945-bda8-8803d08eb16f","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"James Codella, Sergiy Smyrnov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493196687001nAO6","langLocale":"en-US","sessionId":"1774493196687001nAO6","sessionInstanceId":"1774493196687001nAO6","sessionCode":"TT621","sessionCodeNormalized":"EN-US-TT621","title":"Build & modernize multi-tenant AI app data planes for planetary scale","sortTitle":"build & modernize multi-tenant ai app data planes for planetary scale","description":"Most AI-agent demos fail in production because their data platforms were never built for multi-tenant scale. This table talk connects two real journeys: modernizing relational systems to NoSQL, and running large-scale agentic SaaS. We’ll whiteboard migration tradeoffs, tenant-aware modeling, agent memory and retrieval patterns, plus hard lessons learned to ship reliable AI systems at scale today!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196687001nAO6_20260515115325399.avif","speakerIds":["1768245294609build26-1776347675875001DHR6","1768245294609build26-1776352718006001ATTV"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.642Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196687001nAO6_20260515115325399.avif","ogTitle":"Build & modernize multi-tenant AI app data planes for planetary scale","ogDescription":"Most AI-agent demos fail in production because their data platforms were never built for multi-tenant scale. This table talk connects two real journeys: modernizing relational systems to NoSQL, and running large-scale agentic SaaS. We’ll whiteboard migration tradeoffs, tenant-aware modeling, agent memory and retrieval patterns, plus hard lessons learned to ship reliable AI systems at scale today!\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776352718006001ATTV","RegistrantKey":"1768245294609build26-1776352718006001ATTV"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776347675875001DHR6","RegistrantKey":"1768245294609build26-1776347675875001DHR6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"CosmosDB","logicalValue":"CosmosDB"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"22248832-21e7-407f-843e-c55f777893b8","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Moran Assaf, Meng Tang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493197941001nlco","langLocale":"en-US","sessionId":"1774493197941001nlco","sessionInstanceId":"1774493197941001nlco","sessionCode":"TT635","sessionCodeNormalized":"EN-US-TT635","title":"Building Locally with Foundry: Designing AI Agents and Apps","sortTitle":"building locally with foundry: designing ai agents and apps","description":"How do you build and run AI applications locally with Microsoft Foundry? Bring your use cases as we discuss real-world patterns, design tradeoffs, and common challenges from local development to production. This is a conversation-driven session focused on practical decisions you can apply immediately. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197941001nlco_20260602111530255.avif","speakerIds":["1768245294609build26-1778618247524001n8Zk","1768245294609build26-1774383196037001nAji"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.807Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197941001nlco_20260602111530255.avif","ogTitle":"Building Locally with Foundry: Designing AI Agents and Apps","ogDescription":"How do you build and run AI applications locally with Microsoft Foundry? Bring your use cases as we discuss real-world patterns, design tradeoffs, and common challenges from local development to production. This is a conversation-driven session focused on practical decisions you can apply immediately. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774383196037001nAji","RegistrantKey":"1768245294609build26-1774383196037001nAji"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778618247524001n8Zk","RegistrantKey":"1768245294609build26-1778618247524001n8Zk"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"d2fff4ce-b9ef-4c59-80df-1aa7a87d835a","startDateTime":"2026-06-02T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Carlo","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778277073315001SGAI","langLocale":"en-US","sessionId":"1778277073315001SGAI","sessionInstanceId":"1778277073315001SGAI","sessionCode":"TT662B","sessionCodeNormalized":"EN-US-TT662B","title":"Building Strong Community Connections","sortTitle":"building strong community connections","description":"Strong communities don’t happen by accident—they’re built through meaningful connection, shared learning, and consistent engagement. Let's explore how user groups, conferences, online forums, and social media groups can work together to create vibrant, supportive, technical communities where people learn all about from one another, build relationships, and grow their impact. Learn practical ideas for fostering connection, encouraging participation, and strengthening community across both in-person and digital spaces.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778277073315001SGAI_20260514205349732.avif","speakerIds":["1768245294609build26-1776293231988001WPJb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.775Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778277073315001SGAI_20260514205349732.avif","ogTitle":"Building Strong Community Connections","ogDescription":"Strong communities don’t happen by accident—they’re built through meaningful connection, shared learning, and consistent engagement. Let's explore how user groups, conferences, online forums, and social media groups can work together to create vibrant, supportive, technical communities where people learn all about from one another, build relationships, and grow their impact. Learn practical ideas for fostering connection, encouraging participation, and strengthening community across both in-person and digital spaces.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776293231988001WPJb","RegistrantKey":"1768245294609build26-1776293231988001WPJb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"adb21001-e540-454b-a803-3d6486be9391","startDateTime":"2026-06-02T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rie Merritt","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283580001kw2T","langLocale":"en-US","sessionId":"1774493283580001kw2T","sessionInstanceId":"1774493283580001kw2T","sessionCode":"TT662A","sessionCodeNormalized":"EN-US-TT662A","title":"Building Strong Community Connections","sortTitle":"building strong community connections","description":"Strong communities don’t happen by accident—they’re built through meaningful connection, shared learning, and consistent engagement. Let's explore how user groups, conferences, online forums, and social media groups can work together to create vibrant, supportive, technical communities where people learn all about from one another, build relationships, and grow their impact. Learn practical ideas for fostering connection, encouraging participation, and strengthening community across both in-person and digital spaces.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283580001kw2T_20260514205105869.avif","speakerIds":["1768245294609build26-1778277011891001WLB9"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.824Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283580001kw2T_20260514205105869.avif","ogTitle":"Building Strong Community Connections","ogDescription":"Strong communities don’t happen by accident—they’re built through meaningful connection, shared learning, and consistent engagement. Let's explore how user groups, conferences, online forums, and social media groups can work together to create vibrant, supportive, technical communities where people learn all about from one another, build relationships, and grow their impact. Learn practical ideas for fostering connection, encouraging participation, and strengthening community across both in-person and digital spaces.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778277011891001WLB9","RegistrantKey":"1768245294609build26-1778277011891001WLB9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"3682571c-c416-4152-bd7d-7ae68955d6f3","startDateTime":"2026-06-03T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Gresham","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282998001khhy","langLocale":"en-US","sessionId":"1774493282998001khhy","sessionInstanceId":"1774493282998001khhy","sessionCode":"TT658","sessionCodeNormalized":"EN-US-TT658","title":"Building with AI on Systems You Don’t Fully Understand","sortTitle":"building with ai on systems you don’t fully understand","description":"Most teams aren’t building on greenfield; they’re layering AI onto messy, poorly understood systems. What breaks when developers and AI agents lack visibility into system architecture and logic? In this discussion, we’ll compare approaches teams use today: reverse engineering, documentation methods, and queryable system intelligence. We’ll examine tradeoffs in speed, accuracy, and risk, and share ways to make systems more navigable for humans and AI so you can ship faster with fewer surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282998001khhy_20260518110659308.avif","speakerIds":["1768245294609build26-1777732377222001hjKZ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.823Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282998001khhy_20260518110659308.avif","ogTitle":"Building with AI on Systems You Don’t Fully Understand","ogDescription":"Most teams aren’t building on greenfield; they’re layering AI onto messy, poorly understood systems. What breaks when developers and AI agents lack visibility into system architecture and logic? In this discussion, we’ll compare approaches teams use today: reverse engineering, documentation methods, and queryable system intelligence. We’ll examine tradeoffs in speed, accuracy, and risk, and share ways to make systems more navigable for humans and AI so you can ship faster with fewer surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777732377222001hjKZ","RegistrantKey":"1768245294609build26-1777732377222001hjKZ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"a49616b8-510b-4f60-85c4-0b20fe3c0185","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Gresham","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778766235363001E05V","langLocale":"en-US","sessionId":"1778766235363001E05V","sessionInstanceId":"1778766235363001E05V","sessionCode":"TT658-R1","sessionCodeNormalized":"EN-US-TT658-R1","title":"Building with AI on Systems You Don’t Fully Understand","sortTitle":"building with ai on systems you don’t fully understand","description":"Most teams aren’t building on greenfield; they’re layering AI onto messy, poorly understood systems. What breaks when developers and AI agents lack visibility into system architecture and logic? In this discussion, we’ll compare approaches teams use today: reverse engineering, documentation methods, and queryable system intelligence. We’ll examine tradeoffs in speed, accuracy, and risk, and share ways to make systems more navigable for humans and AI so you can ship faster with fewer surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778766235363001E05V_20260518110835326.avif","speakerIds":["1768245294609build26-1777732377222001hjKZ"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.233Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778766235363001E05V_20260518110835326.avif","ogTitle":"Building with AI on Systems You Don’t Fully Understand","ogDescription":"Most teams aren’t building on greenfield; they’re layering AI onto messy, poorly understood systems. What breaks when developers and AI agents lack visibility into system architecture and logic? In this discussion, we’ll compare approaches teams use today: reverse engineering, documentation methods, and queryable system intelligence. We’ll examine tradeoffs in speed, accuracy, and risk, and share ways to make systems more navigable for humans and AI so you can ship faster with fewer surprises.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777732377222001hjKZ","RegistrantKey":"1768245294609build26-1777732377222001hjKZ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"c6726bc8-e523-4d6f-b565-e569cbbd71cc","startDateTime":"2026-06-03T18:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kirupa Chinnathambi, Stuart Schaefer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777930351320001VdN0","langLocale":"en-US","sessionId":"1777930351320001VdN0","sessionInstanceId":"1777930351320001VdN0","sessionCode":"TT645-R1","sessionCodeNormalized":"EN-US-TT645-R1","title":"Claws on Windows: Designing Safe, Bounded Agent Actions","sortTitle":"claws on windows: designing safe, bounded agent actions","description":"Claws touch files, processes, networks, and user state. Poorly designed claws create security risk and brittle behavior. In this interactive table talk, two experts will break down real claw designs used today, including permission scoping, execution boundaries, observability, and rollback. Developers will compare approaches, challenge assumptions, and share what has failed in practice while shaping clearer guidance for building reliable claws on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"18:15 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777930351320001VdN0_20260512031703838.avif","speakerIds":["1768245294609build26-1774284992189001pC5a","1768245294609build26-1776972565872001qoXF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.767Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777930351320001VdN0_20260512031703838.avif","ogTitle":"Claws on Windows: Designing Safe, Bounded Agent Actions","ogDescription":"Claws touch files, processes, networks, and user state. Poorly designed claws create security risk and brittle behavior. In this interactive table talk, two experts will break down real claw designs used today, including permission scoping, execution boundaries, observability, and rollback. Developers will compare approaches, challenge assumptions, and share what has failed in practice while shaping clearer guidance for building reliable claws on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776972565872001qoXF","RegistrantKey":"1768245294609build26-1776972565872001qoXF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284992189001pC5a","RegistrantKey":"1768245294609build26-1774284992189001pC5a"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Claws","logicalValue":"Claws"},{"displayValue":"Openclaw","logicalValue":"Openclaw"}]},{"@search.score":1.0,"documentId":"9cee5e16-df2b-4aab-b089-9fafa4826cb0","startDateTime":"2026-06-02T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kirupa Chinnathambi, Stuart Schaefer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493198891001nuuk","langLocale":"en-US","sessionId":"1774493198891001nuuk","sessionInstanceId":"1774493198891001nuuk","sessionCode":"TT645","sessionCodeNormalized":"EN-US-TT645","title":"Claws on Windows: Designing Safe, Bounded Agent Actions","sortTitle":"claws on windows: designing safe, bounded agent actions","description":"Claws touch files, processes, networks, and user state. Poorly designed claws create security risk and brittle behavior. In this interactive table talk, two experts will break down real claw designs used today, including permission scoping, execution boundaries, observability, and rollback. Developers will compare approaches, challenge assumptions, and share what has failed in practice while shaping clearer guidance for building reliable claws on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198891001nuuk_20260512024057848.avif","speakerIds":["1768245294609build26-1774284992189001pC5a","1768245294609build26-1776972565872001qoXF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.808Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198891001nuuk_20260512024057848.avif","ogTitle":"Claws on Windows: Designing Safe, Bounded Agent Actions","ogDescription":"Claws touch files, processes, networks, and user state. Poorly designed claws create security risk and brittle behavior. In this interactive table talk, two experts will break down real claw designs used today, including permission scoping, execution boundaries, observability, and rollback. Developers will compare approaches, challenge assumptions, and share what has failed in practice while shaping clearer guidance for building reliable claws on Windows.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776972565872001qoXF","RegistrantKey":"1768245294609build26-1776972565872001qoXF"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284992189001pC5a","RegistrantKey":"1768245294609build26-1774284992189001pC5a"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Claws","logicalValue":"Claws"},{"displayValue":"Openclaw","logicalValue":"Openclaw"}]},{"@search.score":1.0,"documentId":"c1e1ab9a-e9bb-4e17-a359-9101c37e5a50","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sunitha Muthukrishna, Ben Zulauf","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493196541001nku3","langLocale":"en-US","sessionId":"1774493196541001nku3","sessionInstanceId":"1774493196541001nku3","sessionCode":"TT620","sessionCodeNormalized":"EN-US-TT620","title":"Coding Agents in the Enterprise: What's Working? What's Not?","sortTitle":"coding agents in the enterprise: what's working? what's not?","description":"With AI-assisted development on the rise, teams are rethinking how they build apps. They're using agents to write code, connect to data, provision services, and set access policies, all within existing compliance requirements. A common thread we're hearing: teams are struggling to unleash coding agents safely and maintainably. In this table talk, we'd love to hear where agents create the most value for your team and what keeps you up at night. We'll share learnings and best practices.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196541001nku3_20260512023913092.avif","speakerIds":["1768245294609build26-1776443914478001L3uX","1768245294609build26-1774284988298001pfqa"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.642Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196541001nku3_20260512023913092.avif","ogTitle":"Coding Agents in the Enterprise: What's Working? What's Not?","ogDescription":"With AI-assisted development on the rise, teams are rethinking how they build apps. They're using agents to write code, connect to data, provision services, and set access policies, all within existing compliance requirements. A common thread we're hearing: teams are struggling to unleash coding agents safely and maintainably. In this table talk, we'd love to hear where agents create the most value for your team and what keeps you up at night. We'll share learnings and best practices.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776443914478001L3uX","RegistrantKey":"1768245294609build26-1776443914478001L3uX"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284988298001pfqa","RegistrantKey":"1768245294609build26-1774284988298001pfqa"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"}]},{"@search.score":1.0,"documentId":"00f81ded-6508-4339-8772-84e8cc2eb81d","startDateTime":"2026-06-03T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Balan Subramanian, Divya Swarnkar","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493198237001np9L","langLocale":"en-US","sessionId":"1774493198237001np9L","sessionInstanceId":"1774493198237001np9L","sessionCode":"TT637","sessionCodeNormalized":"EN-US-TT637","title":"Designing Agentic Workflows: Patterns and Tradeoffs","sortTitle":"designing agentic workflows: patterns and tradeoffs","description":"Agentic workflows combine APIs, AI reasoning, and human decisions across systems—introducing new challenges in orchestration, reliability, and observability at scale. In this table talk, we’ll explore how teams are designing agentic business processes, comparing approaches to combining deterministic workflows and AI-driven decisioning. Join to discuss tradeoffs, real-world challenges, and lessons from moving a single workflow into production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198237001np9L_20260512153128044.avif","speakerIds":["1768245294609build26-1777577534236001U3eH","1768245294609build26-1777934900615001XSkq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.792Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198237001np9L_20260512153128044.avif","ogTitle":"Designing Agentic Workflows: Patterns and Tradeoffs","ogDescription":"Agentic workflows combine APIs, AI reasoning, and human decisions across systems—introducing new challenges in orchestration, reliability, and observability at scale. In this table talk, we’ll explore how teams are designing agentic business processes, comparing approaches to combining deterministic workflows and AI-driven decisioning. Join to discuss tradeoffs, real-world challenges, and lessons from moving a single workflow into production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777577534236001U3eH","RegistrantKey":"1768245294609build26-1777577534236001U3eH"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777934900615001XSkq","RegistrantKey":"1768245294609build26-1777934900615001XSkq"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Enterprise","logicalValue":"Enterprise"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"58b67ca7-57db-423a-9f6f-41ac80c76903","startDateTime":"2026-06-03T18:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jon Hsieh, Amid Tabrizi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282255001kTI8","langLocale":"en-US","sessionId":"1774493282255001kTI8","sessionInstanceId":"1774493282255001kTI8","sessionCode":"TT653","sessionCodeNormalized":"EN-US-TT653","title":"Designing AI Data Platforms on Azure Blob Beyond Retrieval","sortTitle":"designing ai data platforms on azure blob beyond retrieval","description":"You have data in Azure Blob Storage. You add embeddings, connect to search, and ship. It works until you need to change the data itself. Curation, feature engineering, training, search, and analytics quickly spread data across multiple systems. This discussion explores how teams design AI data platforms on Azure beyond retrieval, comparing multi-system setups with approaches that treat Blob as the unified foundation for raw data, embeddings, and features, and the tradeoffs that come with each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"18:15 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282255001kTI8_20260521160822241.avif","speakerIds":["1768245294609build26-1779202833477001OjA8","1768245294609build26-1777930600669001tGiz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.808Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282255001kTI8_20260521160822241.avif","ogTitle":"Designing AI Data Platforms on Azure Blob Beyond Retrieval","ogDescription":"You have data in Azure Blob Storage. You add embeddings, connect to search, and ship. It works until you need to change the data itself. Curation, feature engineering, training, search, and analytics quickly spread data across multiple systems. This discussion explores how teams design AI data platforms on Azure beyond retrieval, comparing multi-system setups with approaches that treat Blob as the unified foundation for raw data, embeddings, and features, and the tradeoffs that come with each.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777930600669001tGiz","RegistrantKey":"1768245294609build26-1777930600669001tGiz"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1779202833477001OjA8","RegistrantKey":"1768245294609build26-1779202833477001OjA8"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Storage","logicalValue":"Storage"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"26ab5fff-f036-42cc-bca0-9d72fa68ddf4","startDateTime":"2026-06-03T18:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Balan Subramanian, Sreekanth Venkata","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493198090001nDrp","langLocale":"en-US","sessionId":"1774493198090001nDrp","sessionInstanceId":"1774493198090001nDrp","sessionCode":"TT636","sessionCodeNormalized":"EN-US-TT636","title":"Designing AI Gateways: Routing, Cost, and Control","sortTitle":"designing ai gateways: routing, cost, and control","description":"AI workloads don’t behave like traditional APIs—they span models, tools, and agents across providers, making routing, cost control, safety, and observability harder in production. In this table talk, we’ll explore how teams are designing AI gateway patterns, comparing approaches for multi-model routing, policy enforcement, and cost management. Join to discuss tradeoffs, production challenges, and what works when scaling AI systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"18:15 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198090001nDrp_20260512153118803.avif","speakerIds":["1768245294609build26-1777577534236001U3eH","1768245294609build26-1778168618206001zt4S"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.807Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493198090001nDrp_20260512153118803.avif","ogTitle":"Designing AI Gateways: Routing, Cost, and Control","ogDescription":"AI workloads don’t behave like traditional APIs—they span models, tools, and agents across providers, making routing, cost control, safety, and observability harder in production. In this table talk, we’ll explore how teams are designing AI gateway patterns, comparing approaches for multi-model routing, policy enforcement, and cost management. Join to discuss tradeoffs, production challenges, and what works when scaling AI systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777577534236001U3eH","RegistrantKey":"1768245294609build26-1777577534236001U3eH"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778168618206001zt4S","RegistrantKey":"1768245294609build26-1778168618206001zt4S"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"9dad8848-41fb-471e-b907-607fce172376","startDateTime":"2026-06-03T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ernesto Cardenas Cangahuala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779812779246001dNbF","langLocale":"en-US","sessionId":"1779812779246001dNbF","sessionInstanceId":"1779812779246001dNbF","sessionCode":"TT679","sessionCodeNormalized":"EN-US-TT679","title":"DevOps adoption: challenges from the trenches","sortTitle":"devops adoption: challenges from the trenches","description":"DevOps adoption is not only about tools. This interactive Table Talk explores real-world lessons on technical blockers, cultural resistance, and hidden complexity that emerges when teams adopt DevOps practices. We will compare practical patterns, common mistakes, and how platform engineering and AI are influencing the next stage of DevOps, so attendees leave with useful ideas for their own organizations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286155001k4vR_20260512024349370.avif","speakerIds":["1768245294609build26-1776296349532001Li4L"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.238Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286155001k4vR_20260512024349370.avif","ogTitle":"DevOps adoption: challenges from the trenches","ogDescription":"DevOps adoption is not only about tools. This interactive Table Talk explores real-world lessons on technical blockers, cultural resistance, and hidden complexity that emerges when teams adopt DevOps practices. We will compare practical patterns, common mistakes, and how platform engineering and AI are influencing the next stage of DevOps, so attendees leave with useful ideas for their own organizations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296349532001Li4L","RegistrantKey":"1768245294609build26-1776296349532001Li4L"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"52594e30-690e-4aff-aba5-61d122d1c74e","startDateTime":"2026-06-03T18:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ernesto Cardenas Cangahuala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493286155001k4vR","langLocale":"en-US","sessionId":"1774493286155001k4vR","sessionInstanceId":"1774493286155001k4vR","sessionCode":"TT679-R1","sessionCodeNormalized":"EN-US-TT679-R1","title":"DevOps adoption: challenges from the trenches","sortTitle":"devops adoption: challenges from the trenches","description":"DevOps adoption is not only about tools. This interactive Table Talk explores real-world lessons on technical blockers, cultural resistance, and hidden complexity that emerges when teams adopt DevOps practices. We will compare practical patterns, common mistakes, and how platform engineering and AI are influencing the next stage of DevOps, so attendees leave with useful ideas for their own organizations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"18:15 - 19:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T19:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812779246001dNbF_20260527112631756.avif","speakerIds":["1768245294609build26-1776296349532001Li4L"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.84Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812779246001dNbF_20260527112631756.avif","ogTitle":"DevOps adoption: challenges from the trenches","ogDescription":"DevOps adoption is not only about tools. This interactive Table Talk explores real-world lessons on technical blockers, cultural resistance, and hidden complexity that emerges when teams adopt DevOps practices. We will compare practical patterns, common mistakes, and how platform engineering and AI are influencing the next stage of DevOps, so attendees leave with useful ideas for their own organizations.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296349532001Li4L","RegistrantKey":"1768245294609build26-1776296349532001Li4L"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"08436ef9-574e-4137-b906-52fe90341bcd","startDateTime":"2026-06-03T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nelly Delgado","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779458697650001Y65q","langLocale":"en-US","sessionId":"1779458697650001Y65q","sessionInstanceId":"1779458697650001Y65q","sessionCode":"TT665B","sessionCodeNormalized":"EN-US-TT665B","title":"Experimentation to Scale: Navigating Real-World AI Transformation","sortTitle":"experimentation to scale: navigating real-world ai transformation","description":"AI is everywhere—but scaling it across organizations is still hard.\n\nIn this roundtable, we’ll move beyond hype to discuss the real challenges developers face when helping organizations adopt AI—from integrating into existing architectures and workflows to managing risk, governance, and change.\n\nJoin peers to share lessons learned, explore what’s working (and what isn’t), and uncover practical approaches to:\n•\tMoving from pilots to production\n•\tEnabling responsible and secure AI adoption\n•\tDriving alignment between engineering, business, and customers\n\nThis is an interactive session designed for developers actively shaping how AI is implemented—both inside their organizations and the customers they serve.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779458697650001Y65q_20260526104649999.avif","speakerIds":["1768245294609build26-1770330200998001uWJm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.236Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779458697650001Y65q_20260526104649999.avif","ogTitle":"Experimentation to Scale: Navigating Real-World AI Transformation","ogDescription":"AI is everywhere—but scaling it across organizations is still hard.\n\nIn this roundtable, we’ll move beyond hype to discuss the real challenges developers face when helping organizations adopt AI—from integrating into existing architectures and workflows to managing risk, governance, and change.\n\nJoin peers to share lessons learned, explore what’s working (and what isn’t), and uncover practical approaches to:\n•\tMoving from pilots to production\n•\tEnabling responsible and secure AI adoption\n•\tDriving alignment between engineering, business, and customers\n\nThis is an interactive session designed for developers actively shaping how AI is implemented—both inside their organizations and the customers they serve.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770330200998001uWJm","RegistrantKey":"1768245294609build26-1770330200998001uWJm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"b1c248c9-5791-43b3-bf2a-932056e70340","startDateTime":"2026-06-03T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Marisela Cerda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493284036001kHbn","langLocale":"en-US","sessionId":"1774493284036001kHbn","sessionInstanceId":"1774493284036001kHbn","sessionCode":"TT665A","sessionCodeNormalized":"EN-US-TT665A","title":"Experimentation to Scale: Navigating Real-World AI Transformation","sortTitle":"experimentation to scale: navigating real-world ai transformation","description":"AI is everywhere—but scaling it across organizations is still hard.\n\nIn this roundtable, we’ll move beyond hype to discuss the real challenges developers face when helping organizations adopt AI—from integrating into existing architectures and workflows to managing risk, governance, and change.\n\nJoin peers to share lessons learned, explore what’s working (and what isn’t), and uncover practical approaches to:\n•\tMoving from pilots to production\n•\tEnabling responsible and secure AI adoption\n•\tDriving alignment between engineering, business, and customers\n\nThis is an interactive session designed for developers actively shaping how AI is implemented—both inside their organizations and the customers they serve.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284036001kHbn_20260526104914067.avif","speakerIds":["1768245294609build26-1770320163509001ehOc"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.839Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284036001kHbn_20260526104914067.avif","ogTitle":"Experimentation to Scale: Navigating Real-World AI Transformation","ogDescription":"AI is everywhere—but scaling it across organizations is still hard.\n\nIn this roundtable, we’ll move beyond hype to discuss the real challenges developers face when helping organizations adopt AI—from integrating into existing architectures and workflows to managing risk, governance, and change.\n\nJoin peers to share lessons learned, explore what’s working (and what isn’t), and uncover practical approaches to:\n•\tMoving from pilots to production\n•\tEnabling responsible and secure AI adoption\n•\tDriving alignment between engineering, business, and customers\n\nThis is an interactive session designed for developers actively shaping how AI is implemented—both inside their organizations and the customers they serve.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770320163509001ehOc","RegistrantKey":"1768245294609build26-1770320163509001ehOc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"9a1418a3-2327-495e-b283-b04f5c4d931c","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ander Steele","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493281962001km5f","langLocale":"en-US","sessionId":"1774493281962001km5f","sessionInstanceId":"1774493281962001km5f","sessionCode":"TT651","sessionCodeNormalized":"EN-US-TT651","title":"Fake It or Mask It: Choosing the Right Data for Agent Simulation Environments","sortTitle":"fake it or mask it: choosing the right data for agent simulation environments","description":"Agents need realistic worlds to train in, but the data behind them is rarely safe to use raw. Two paths dominate: synthetic data that mirrors production, or de-identified real data that keeps its messiness without exposing real people. Each makes different tradeoffs on fidelity, edge cases, governance, and generalization. Let's compare notes on where each shine, where it breaks, and where teams blend the two.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493281962001km5f_20260514205039102.avif","speakerIds":["1768245294609build26-1778161701758001C6sB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.809Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493281962001km5f_20260514205039102.avif","ogTitle":"Fake It or Mask It: Choosing the Right Data for Agent Simulation Environments","ogDescription":"Agents need realistic worlds to train in, but the data behind them is rarely safe to use raw. Two paths dominate: synthetic data that mirrors production, or de-identified real data that keeps its messiness without exposing real people. Each makes different tradeoffs on fidelity, edge cases, governance, and generalization. Let's compare notes on where each shine, where it breaks, and where teams blend the two.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778161701758001C6sB","RegistrantKey":"1768245294609build26-1778161701758001C6sB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"2fd82234-6d9a-4ef1-90d2-0052e34cf3af","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Matt Campbell","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282701001kcA4","langLocale":"en-US","sessionId":"1774493282701001kcA4","sessionInstanceId":"1774493282701001kcA4","sessionCode":"TT656","sessionCodeNormalized":"EN-US-TT656","title":"From Locked-In to Liquid: Modernizing .NET Before the November 2026 EOL","sortTitle":"from locked-in to liquid: modernizing .net before the november 2026 eol","description":".NET 8 and 9 both hit end of life on November 10, 2026. With hundreds of repos to upgrade, manual migration won't cut it—and pointing agents at the problem without structure burns tokens and produces inconsistent results. Let's talk about how agents, given the right deterministic tools, can move your .NET portfolio from rigid and locked-in to liquid—continuously modern, with swappable components and routine upgrades instead of fire drills. Bring your migration questions and stack horror stories.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282701001kcA4_20260512024214239.avif","speakerIds":["1768245294609build26-1778001554148001dPcn"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.809Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282701001kcA4_20260512024214239.avif","ogTitle":"From Locked-In to Liquid: Modernizing .NET Before the November 2026 EOL","ogDescription":".NET 8 and 9 both hit end of life on November 10, 2026. With hundreds of repos to upgrade, manual migration won't cut it—and pointing agents at the problem without structure burns tokens and produces inconsistent results. Let's talk about how agents, given the right deterministic tools, can move your .NET portfolio from rigid and locked-in to liquid—continuously modern, with swappable components and routine upgrades instead of fire drills. Bring your migration questions and stack horror stories.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1778001554148001dPcn","RegistrantKey":"1768245294609build26-1778001554148001dPcn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"DevTools","logicalValue":"DevTools"}]},{"@search.score":1.0,"documentId":"c79e8e8d-500a-4949-a1f6-7d00c9ed6d89","startDateTime":"2026-06-03T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Stephen Simon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283726001kA0H","langLocale":"en-US","sessionId":"1774493283726001kA0H","sessionInstanceId":"1774493283726001kA0H","sessionCode":"TT663A","sessionCodeNormalized":"EN-US-TT663A","title":"Global AI Community – Connect, Learn & Share Worldwide","sortTitle":"global ai community – connect, learn & share worldwide","description":"Curious about the Global AI Community? Join this casual meetup to connect with AI enthusiasts worldwide. Learn how the Global AI Community helps you share ideas, build skills, and access resources—whether you're just starting out or deep in AI development. Meet Microsoft experts and community organizers, ask questions, and discover how to get involved in local meetups and global events. Stop by to ask questions and discover how easy it is to get involved and grow your AI network.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"16:15 - 17:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283726001kA0H_20260512032151424.avif","speakerIds":["1768245294609build26-1774323608465001D4tS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.824Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283726001kA0H_20260512032151424.avif","ogTitle":"Global AI Community – Connect, Learn & Share Worldwide","ogDescription":"Curious about the Global AI Community? Join this casual meetup to connect with AI enthusiasts worldwide. Learn how the Global AI Community helps you share ideas, build skills, and access resources—whether you're just starting out or deep in AI development. Meet Microsoft experts and community organizers, ask questions, and discover how to get involved in local meetups and global events. Stop by to ask questions and discover how easy it is to get involved and grow your AI network.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774323608465001D4tS","RegistrantKey":"1768245294609build26-1774323608465001D4tS"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"f4601dcb-2ee0-49dd-8885-9a09e2329073","startDateTime":"2026-06-03T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roelant Dieben","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778597035249001MtSd","langLocale":"en-US","sessionId":"1778597035249001MtSd","sessionInstanceId":"1778597035249001MtSd","sessionCode":"TT663B","sessionCodeNormalized":"EN-US-TT663B","title":"Global AI Community – Connect, Learn & Share Worldwide","sortTitle":"global ai community – connect, learn & share worldwide","description":"Curious about the Global AI Community? Join this casual meetup to connect with AI enthusiasts worldwide. Learn how the Global AI Community helps you share ideas, build skills, and access resources—whether you're just starting out or deep in AI development. Meet Microsoft experts and community organizers, ask questions, and discover how to get involved in local meetups and global events. Stop by to ask questions and discover how easy it is to get involved and grow your AI network.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"16:15 - 17:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283726001kA0H_20260512032151424.avif","speakerIds":["1768245294609build26-1770888477835001mAZA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.231Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283726001kA0H_20260512032151424.avif","ogTitle":"Global AI Community – Connect, Learn & Share Worldwide","ogDescription":"Curious about the Global AI Community? Join this casual meetup to connect with AI enthusiasts worldwide. Learn how the Global AI Community helps you share ideas, build skills, and access resources—whether you're just starting out or deep in AI development. Meet Microsoft experts and community organizers, ask questions, and discover how to get involved in local meetups and global events. Stop by to ask questions and discover how easy it is to get involved and grow your AI network.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770888477835001mAZA","RegistrantKey":"1768245294609build26-1770888477835001mAZA"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"51bc3f49-89f8-4931-9c32-f069c70b9e6c","startDateTime":"2026-06-03T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Brittney Smith","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778597370858001wW3a","langLocale":"en-US","sessionId":"1778597370858001wW3a","sessionInstanceId":"1778597370858001wW3a","sessionCode":"TT667B","sessionCodeNormalized":"EN-US-TT667B","title":"Growing Local Developer Communities with Azure Tech Groups","sortTitle":"growing local developer communities with azure tech groups","description":"Are you a community leader or looking for more ways to get involved with your local developer community? Interested in creating your own community? Join Azure Tech Group (ATG) leaders to learn how ATG connects developers, cloud engineers, and tech enthusiasts through local meetups, networking, and collaboration. Discover how to start and grow a community, connect with Microsoft experts, share real-world AI and Azure experiences, and engage with the broader Azure technical ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284329001kLxr_20260512032229987.avif","speakerIds":["1768245294609build26-1772126555167001lJnX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.231Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284329001kLxr_20260512032229987.avif","ogTitle":"Growing Local Developer Communities with Azure Tech Groups","ogDescription":"Are you a community leader or looking for more ways to get involved with your local developer community? Interested in creating your own community? Join Azure Tech Group (ATG) leaders to learn how ATG connects developers, cloud engineers, and tech enthusiasts through local meetups, networking, and collaboration. Discover how to start and grow a community, connect with Microsoft experts, share real-world AI and Azure experiences, and engage with the broader Azure technical ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1772126555167001lJnX","RegistrantKey":"1768245294609build26-1772126555167001lJnX"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"c4d6b983-a22d-431b-af92-2989ea64a56e","startDateTime":"2026-06-03T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Galimah Baysah","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493284329001kLxr","langLocale":"en-US","sessionId":"1774493284329001kLxr","sessionInstanceId":"1774493284329001kLxr","sessionCode":"TT667A","sessionCodeNormalized":"EN-US-TT667A","title":"Growing Local Developer Communities with Azure Tech Groups","sortTitle":"growing local developer communities with azure tech groups","description":"Are you a community leader or looking for more ways to get involved with your local developer community? Interested in creating your own community? Join Azure Tech Group (ATG) leaders to learn how ATG connects developers, cloud engineers, and tech enthusiasts through local meetups, networking, and collaboration. Discover how to start and grow a community, connect with Microsoft experts, share real-world AI and Azure experiences, and engage with the broader Azure technical ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284329001kLxr_20260512032229987.avif","speakerIds":["1768245294609build26-1759165935261001FbF9"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.84Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493284329001kLxr_20260512032229987.avif","ogTitle":"Growing Local Developer Communities with Azure Tech Groups","ogDescription":"Are you a community leader or looking for more ways to get involved with your local developer community? Interested in creating your own community? Join Azure Tech Group (ATG) leaders to learn how ATG connects developers, cloud engineers, and tech enthusiasts through local meetups, networking, and collaboration. Discover how to start and grow a community, connect with Microsoft experts, share real-world AI and Azure experiences, and engage with the broader Azure technical ecosystem.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1759165935261001FbF9","RegistrantKey":"1768245294609build26-1759165935261001FbF9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"5419fa97-8885-4863-8767-11e958f2d487","startDateTime":"2026-06-03T21:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778881781483001Ow6f","langLocale":"en-US","sessionId":"1778881781483001Ow6f","sessionInstanceId":"1778881781483001Ow6f","sessionCode":"TT697","sessionCodeNormalized":"EN-US-TT697","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:00 - 21:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:30:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881781483001Ow6f_20260518110441644.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.234Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881781483001Ow6f_20260518110441644.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"b0a3fb42-203f-4d39-abb2-69e365193b09","startDateTime":"2026-06-02T21:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778690980593001E0XF","langLocale":"en-US","sessionId":"1778690980593001E0XF","sessionInstanceId":"1778690980593001E0XF","sessionCode":"TT694","sessionCodeNormalized":"EN-US-TT694","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:30 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778690980593001E0XF_20260514112335558.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.232Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778690980593001E0XF_20260514112335558.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"957a337f-59ed-4e66-b741-c7cb66d901da","startDateTime":"2026-06-03T00:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778881821143001TOg0","langLocale":"en-US","sessionId":"1778881821143001TOg0","sessionInstanceId":"1778881821143001TOg0","sessionCode":"TT696","sessionCodeNormalized":"EN-US-TT696","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:00 - 00:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:30:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881821143001TOg0_20260518110451924.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.234Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881821143001TOg0_20260518110451924.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"6b788a45-bbc1-4fa8-9213-4d596bf7dec6","startDateTime":"2026-06-04T00:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778881760415001TcTW","langLocale":"en-US","sessionId":"1778881760415001TcTW","sessionInstanceId":"1778881760415001TcTW","sessionCode":"TT699","sessionCodeNormalized":"EN-US-TT699","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:30 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881760415001TcTW_20260518110425672.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.234Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881760415001TcTW_20260518110425672.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"fad2d69e-b1a6-4813-8ddb-0d88f1b4a189","startDateTime":"2026-06-02T23:00:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778881626547001gHKk","langLocale":"en-US","sessionId":"1778881626547001gHKk","sessionInstanceId":"1778881626547001gHKk","sessionCode":"TT695","sessionCodeNormalized":"EN-US-TT695","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:00 - 23:30","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:30:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881626547001gHKk_20260518110415550.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.234Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881626547001gHKk_20260518110415550.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"6ec6f564-ee87-4afe-abaf-56b002243925","startDateTime":"2026-06-03T22:30:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778881761059001Txfg","langLocale":"en-US","sessionId":"1778881761059001Txfg","sessionInstanceId":"1778881761059001Txfg","sessionCode":"TT698","sessionCodeNormalized":"EN-US-TT698","title":"Ignite Research Focus Group","sortTitle":"ignite research focus group","description":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:30 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":30,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881761059001Txfg_20260518110433870.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.234Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778881761059001Txfg_20260518110433870.avif","ogTitle":"Ignite Research Focus Group","ogDescription":"Exclusive event research focus group to share specific feedback about event experience.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 1, Across from Lab 3","logicalValue":"Gateway Pavilion, Level 1, Across from Lab 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"70d26c06-b407-4a8f-95db-855e71ef05c9","startDateTime":"2026-06-03T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nish Anil, Hazem El-Hammamy, Ryan Frazier","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493197153001nhsC","langLocale":"en-US","sessionId":"1774493197153001nhsC","sessionInstanceId":"1774493197153001nhsC","sessionCode":"TT624","sessionCodeNormalized":"EN-US-TT624","title":"Modernizing at scale with agents: Patterns, pitfalls, and tradeoffs","sortTitle":"modernizing at scale with agents: patterns, pitfalls, and tradeoffs","description":"Agents promise to compress years of app modernization into quarters, but speed isn’t the only goal. Enterprises still need to prioritize the right apps, define agent autonomy, and balance velocity with risk and cost. In this discussion, we’ll examine how teams are sequencing portfolios, designing agent workflows, and enforcing guardrails, along with lessons on where agents deliver real value and where they don’t yet.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197153001nhsC_20260512024002757.avif","speakerIds":["1768245294609build26-1770924720596001wFnD","1768245294609build26-1777576114739001uRBc","1768245294609build26-1775581595084001poYl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.642Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197153001nhsC_20260512024002757.avif","ogTitle":"Modernizing at scale with agents: Patterns, pitfalls, and tradeoffs","ogDescription":"Agents promise to compress years of app modernization into quarters, but speed isn’t the only goal. Enterprises still need to prioritize the right apps, define agent autonomy, and balance velocity with risk and cost. In this discussion, we’ll examine how teams are sequencing portfolios, designing agent workflows, and enforcing guardrails, along with lessons on where agents deliver real value and where they don’t yet.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775581595084001poYl","RegistrantKey":"1768245294609build26-1775581595084001poYl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770924720596001wFnD","RegistrantKey":"1768245294609build26-1770924720596001wFnD"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777576114739001uRBc","RegistrantKey":"1768245294609build26-1777576114739001uRBc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"a91546e4-1332-49d3-a86f-daf24d1ea33f","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nish Anil, Hazem El-Hammamy, Ryan Frazier","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779219534306001JgPb","langLocale":"en-US","sessionId":"1779219534306001JgPb","sessionInstanceId":"1779219534306001JgPb","sessionCode":"TT607","sessionCodeNormalized":"EN-US-TT607","title":"Modernizing at scale with agents: Patterns, pitfalls, and tradeoffs","sortTitle":"modernizing at scale with agents: patterns, pitfalls, and tradeoffs","description":"Agents promise to compress years of app modernization into quarters, but speed isn’t the only goal. Enterprises still need to prioritize the right apps, define agent autonomy, and balance velocity with risk and cost. In this discussion, we’ll examine how teams are sequencing portfolios, designing agent workflows, and enforcing guardrails, along with lessons on where agents deliver real value and where they don’t yet.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779219534306001JgPb_20260520160334623.avif","speakerIds":["1768245294609build26-1770924720596001wFnD","1768245294609build26-1777576114739001uRBc","1768245294609build26-1775581595084001poYl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.235Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779219534306001JgPb_20260520160334623.avif","ogTitle":"Modernizing at scale with agents: Patterns, pitfalls, and tradeoffs","ogDescription":"Agents promise to compress years of app modernization into quarters, but speed isn’t the only goal. Enterprises still need to prioritize the right apps, define agent autonomy, and balance velocity with risk and cost. In this discussion, we’ll examine how teams are sequencing portfolios, designing agent workflows, and enforcing guardrails, along with lessons on where agents deliver real value and where they don’t yet.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775581595084001poYl","RegistrantKey":"1768245294609build26-1775581595084001poYl"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770924720596001wFnD","RegistrantKey":"1768245294609build26-1770924720596001wFnD"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777576114739001uRBc","RegistrantKey":"1768245294609build26-1777576114739001uRBc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"0e7dad61-bcd3-476b-ac5f-604bd055d1f9","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ofek Ben Eliezer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779813183164001QgWG","langLocale":"en-US","sessionId":"1779813183164001QgWG","sessionInstanceId":"1779813183164001QgWG","sessionCode":"TT676-R1","sessionCodeNormalized":"EN-US-TT676-R1","title":"Multi-Agent Systems: When (and When Not) to Use Them","sortTitle":"multi-agent systems: when (and when not) to use them","description":"Are multi-agent systems a breakthrough or architectural debt? While the hype is high, many teams over-engineer their AI stacks, leading to increased cost, latency, and complexity. This Table Talk focuses on when multi-agent architectures add real value and when simpler patterns are the better choice. We’ll explore trade-offs between RAG and multi-agent designs and discuss how to maintain control in production systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779813183164001QgWG_20260527112650756.avif","speakerIds":["1768245294609build26-1776296228053001Ld8t"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.238Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779813183164001QgWG_20260527112650756.avif","ogTitle":"Multi-Agent Systems: When (and When Not) to Use Them","ogDescription":"Are multi-agent systems a breakthrough or architectural debt? While the hype is high, many teams over-engineer their AI stacks, leading to increased cost, latency, and complexity. This Table Talk focuses on when multi-agent architectures add real value and when simpler patterns are the better choice. We’ll explore trade-offs between RAG and multi-agent designs and discuss how to maintain control in production systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296228053001Ld8t","RegistrantKey":"1768245294609build26-1776296228053001Ld8t"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"443a834f-699e-427a-8096-f20b8e3719cd","startDateTime":"2026-06-03T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ofek Ben Eliezer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493285670001kzLw","langLocale":"en-US","sessionId":"1774493285670001kzLw","sessionInstanceId":"1774493285670001kzLw","sessionCode":"TT676","sessionCodeNormalized":"EN-US-TT676","title":"Multi-Agent Systems: When (and When Not) to Use Them","sortTitle":"multi-agent systems: when (and when not) to use them","description":"Are multi-agent systems a breakthrough or architectural debt? While the hype is high, many teams over-engineer their AI stacks, leading to increased cost, latency, and complexity. This Table Talk focuses on when multi-agent architectures add real value and when simpler patterns are the better choice. We’ll explore trade-offs between RAG and multi-agent designs and discuss how to maintain control in production systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285670001kzLw_20260512024314451.avif","speakerIds":["1768245294609build26-1776296228053001Ld8t"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.84Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285670001kzLw_20260512024314451.avif","ogTitle":"Multi-Agent Systems: When (and When Not) to Use Them","ogDescription":"Are multi-agent systems a breakthrough or architectural debt? While the hype is high, many teams over-engineer their AI stacks, leading to increased cost, latency, and complexity. This Table Talk focuses on when multi-agent architectures add real value and when simpler patterns are the better choice. We’ll explore trade-offs between RAG and multi-agent designs and discuss how to maintain control in production systems.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296228053001Ld8t","RegistrantKey":"1768245294609build26-1776296228053001Ld8t"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"2b306b23-9b86-457f-b049-ae97651428cd","startDateTime":"2026-06-02T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Arrizza","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282553001kOVC","langLocale":"en-US","sessionId":"1774493282553001kOVC","sessionInstanceId":"1774493282553001kOVC","sessionCode":"TT655","sessionCodeNormalized":"EN-US-TT655","title":"Multimodal data: Architecting pipelines that don’t break at scale","sortTitle":"multimodal data: architecting pipelines that don’t break at scale","description":"Multimodal AI unlocks 90% of the world's data: video, audio, PDFs, sensor streams, etc. But processing it breaks traditional batch pipelines and leaves GPUs idle 50%+ of the time.\n\nIn this session, you'll learn:\n-Why multimodal pipelines break traditional batch architectures\n-How to eliminate CPU to GPU I/O bottlenecks\n-How Anyscale built a streaming CPU+GPU engine with Ray Data\n-How AV customers scale data-intensive workloads on Azure\n\nFor CV researchers, ML engineers, and physical AI teams.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282553001kOVC_20260512024158760.avif","speakerIds":["1768245294609build26-1777486041744001HXer"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.809Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282553001kOVC_20260512024158760.avif","ogTitle":"Multimodal data: Architecting pipelines that don’t break at scale","ogDescription":"Multimodal AI unlocks 90% of the world's data: video, audio, PDFs, sensor streams, etc. But processing it breaks traditional batch pipelines and leaves GPUs idle 50%+ of the time.\n\nIn this session, you'll learn:\n-Why multimodal pipelines break traditional batch architectures\n-How to eliminate CPU to GPU I/O bottlenecks\n-How Anyscale built a streaming CPU+GPU engine with Ray Data\n-How AV customers scale data-intensive workloads on Azure\n\nFor CV researchers, ML engineers, and physical AI teams.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777486041744001HXer","RegistrantKey":"1768245294609build26-1777486041744001HXer"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"b9389c92-4e8a-4802-9b08-44f6b95ec873","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gilad Turbahn","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283143001kiKi","langLocale":"en-US","sessionId":"1774493283143001kiKi","sessionInstanceId":"1774493283143001kiKi","sessionCode":"TT659","sessionCodeNormalized":"EN-US-TT659","title":"Outcome-maxxing, not token-maxxing: what \"good\" looks like in AI-authored code","sortTitle":"outcome-maxxing, not token-maxxing: what \"good\" looks like in ai-authored code","description":"AI is now the primary author in most codebases. Throughput is climbing. Bugs, incidents, churn, and review time are climbing faster. Developers are optimizing for tokens accepted, not code that survives. A working session, not a panel. We'll compare how teams redefine \"good\" in AI-authored code: where to set the bar at authorship vs. review, which guardrails earn their friction, and what to track when dashboards say faster but the team feels slower. Bring an argument you're tired of having.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283143001kiKi_20260528122446808.avif","speakerIds":["1768245294609build26-1777732598190001eWgH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.823Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283143001kiKi_20260528122446808.avif","ogTitle":"Outcome-maxxing, not token-maxxing: what \"good\" looks like in AI-authored code","ogDescription":"AI is now the primary author in most codebases. Throughput is climbing. Bugs, incidents, churn, and review time are climbing faster. Developers are optimizing for tokens accepted, not code that survives. A working session, not a panel. We'll compare how teams redefine \"good\" in AI-authored code: where to set the bar at authorship vs. review, which guardrails earn their friction, and what to track when dashboards say faster but the team feels slower. Bring an argument you're tired of having.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777732598190001eWgH","RegistrantKey":"1768245294609build26-1777732598190001eWgH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"4bc351f7-38f2-42b8-867e-409f2506615c","startDateTime":"2026-06-03T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gilad Turbahn","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778766144222001ffwm","langLocale":"en-US","sessionId":"1778766144222001ffwm","sessionInstanceId":"1778766144222001ffwm","sessionCode":"TT659-R1","sessionCodeNormalized":"EN-US-TT659-R1","title":"Outcome-maxxing, not token-maxxing: what \"good\" looks like in AI-authored code","sortTitle":"outcome-maxxing, not token-maxxing: what \"good\" looks like in ai-authored code","description":"AI is now the primary author in most codebases. Throughput is climbing. Bugs, incidents, churn, and review time are climbing faster. Developers are optimizing for tokens accepted, not code that survives. A working session, not a panel. We'll compare how teams redefine \"good\" in AI-authored code: where to set the bar at authorship vs. review, which guardrails earn their friction, and what to track when dashboards say faster but the team feels slower. Bring an argument you're tired of having.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778766144222001ffwm_20260528123115488.avif","speakerIds":["1768245294609build26-1777732598190001eWgH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.233Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778766144222001ffwm_20260528123115488.avif","ogTitle":"Outcome-maxxing, not token-maxxing: what \"good\" looks like in AI-authored code","ogDescription":"AI is now the primary author in most codebases. Throughput is climbing. Bugs, incidents, churn, and review time are climbing faster. Developers are optimizing for tokens accepted, not code that survives. A working session, not a panel. We'll compare how teams redefine \"good\" in AI-authored code: where to set the bar at authorship vs. review, which guardrails earn their friction, and what to track when dashboards say faster but the team feels slower. Bring an argument you're tired of having.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777732598190001eWgH","RegistrantKey":"1768245294609build26-1777732598190001eWgH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"39ce4b26-2755-4062-b313-9a56a4e88d0f","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nate Barbettini","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778765922248001uBBq","langLocale":"en-US","sessionId":"1778765922248001uBBq","sessionInstanceId":"1778765922248001uBBq","sessionCode":"TT650-R1","sessionCodeNormalized":"EN-US-TT650-R1","title":"Production MCP Servers: Auth, Tools, and the Patterns That Hold Up","sortTitle":"production mcp servers: auth, tools, and the patterns that hold up","description":"Custom agents and platforms like Copilot are only as useful as the actions they can take in the systems that matter. That requires production-grade MCP: secure auth, reliable tools, and the governance the spec leaves to you. We'll cover what holds up under real workloads and the patterns that don't. Bring the problems you're stuck on.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778765922248001uBBq_20260518110817055.avif","speakerIds":["1768245294609build26-1775489841453001uOv5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.233Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778765922248001uBBq_20260518110817055.avif","ogTitle":"Production MCP Servers: Auth, Tools, and the Patterns That Hold Up","ogDescription":"Custom agents and platforms like Copilot are only as useful as the actions they can take in the systems that matter. That requires production-grade MCP: secure auth, reliable tools, and the governance the spec leaves to you. We'll cover what holds up under real workloads and the patterns that don't. Bring the problems you're stuck on.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"TypeScript","logicalValue":"TypeScript"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775489841453001uOv5","RegistrantKey":"1768245294609build26-1775489841453001uOv5"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"6a8186fe-86df-4ee0-a196-9c74f673d04c","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nate Barbettini","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493281790001kBcD","langLocale":"en-US","sessionId":"1774493281790001kBcD","sessionInstanceId":"1774493281790001kBcD","sessionCode":"TT650","sessionCodeNormalized":"EN-US-TT650","title":"Production MCP Servers: Auth, Tools, and the Patterns That Hold Up","sortTitle":"production mcp servers: auth, tools, and the patterns that hold up","description":"Custom agents and platforms like Copilot are only as useful as the actions they can take in the systems that matter. That requires production-grade MCP: secure auth, reliable tools, and the governance the spec leaves to you. We'll cover what holds up under real workloads and the patterns that don't. Bring the problems you're stuck on.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493281790001kBcD_20260518110639394.avif","speakerIds":["1768245294609build26-1775489841453001uOv5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.808Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493281790001kBcD_20260518110639394.avif","ogTitle":"Production MCP Servers: Auth, Tools, and the Patterns That Hold Up","ogDescription":"Custom agents and platforms like Copilot are only as useful as the actions they can take in the systems that matter. That requires production-grade MCP: secure auth, reliable tools, and the governance the spec leaves to you. We'll cover what holds up under real workloads and the patterns that don't. Bring the problems you're stuck on.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"TypeScript","logicalValue":"TypeScript"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1775489841453001uOv5","RegistrantKey":"1768245294609build26-1775489841453001uOv5"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"b398e687-3f7b-4588-a225-da8ecbba25c5","startDateTime":"2026-06-03T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Egina Malaj","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493285518001k21x","langLocale":"en-US","sessionId":"1774493285518001k21x","sessionInstanceId":"1774493285518001k21x","sessionCode":"TT675","sessionCodeNormalized":"EN-US-TT675","title":"Rethinking Retrieval with Foundry IQ","sortTitle":"rethinking retrieval with foundry iq","description":"Most teams build retrieval the same way: custom pipelines on top of search. Foundry IQ aims to improve and changes that. This discussion explores traditional vs agentic retrieval, showing where accuracy improves and where latency tradeoffs appear. We’ll walk through real use cases, built-in multi-retrieval capabilities, and how Foundry IQ integrates with Azure AI Search and other IQ layers. The focus is on when to use it, what it replaces, and how to adopt it without rebuilding everything.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285518001k21x_20260512024302879.avif","speakerIds":["1768245294609build26-1776296191511001xsDz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.84Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493285518001k21x_20260512024302879.avif","ogTitle":"Rethinking Retrieval with Foundry IQ","ogDescription":"Most teams build retrieval the same way: custom pipelines on top of search. Foundry IQ aims to improve and changes that. This discussion explores traditional vs agentic retrieval, showing where accuracy improves and where latency tradeoffs appear. We’ll walk through real use cases, built-in multi-retrieval capabilities, and how Foundry IQ integrates with Azure AI Search and other IQ layers. The focus is on when to use it, what it replaces, and how to adopt it without rebuilding everything.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296191511001xsDz","RegistrantKey":"1768245294609build26-1776296191511001xsDz"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"1ce19049-49ca-4d28-a93d-fd4c968f7f76","startDateTime":"2026-06-04T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Lachie Evenson, Subhav Mital","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778021428918001tJ7R","langLocale":"en-US","sessionId":"1778021428918001tJ7R","sessionInstanceId":"1778021428918001tJ7R","sessionCode":"TT623-R1","sessionCodeNormalized":"EN-US-TT623-R1","title":"Scale Python AI Faster with Anyscale on Azure","sortTitle":"scale python ai faster with anyscale on azure","description":"As Python becomes central to modern AI, many teams struggle to scale from experiments to production. Join this table talk to learn how Anyscale on Azure helps teams move from notebooks to large-scale, distributed AI workloads using open source tools—without managing infrastructure. Discover how managed Ray accelerates training, inference, and data processing, while Azure delivers the security, scale, and reliability needed for enterprise AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-04T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778021428918001tJ7R_20260512031715615.avif","speakerIds":["1768245294609build26-1774284996815001pSYL","1768245294609build26-1776963613993001Zbnm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.768Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778021428918001tJ7R_20260512031715615.avif","ogTitle":"Scale Python AI Faster with Anyscale on Azure","ogDescription":"As Python becomes central to modern AI, many teams struggle to scale from experiments to production. Join this table talk to learn how Anyscale on Azure helps teams move from notebooks to large-scale, distributed AI workloads using open source tools—without managing infrastructure. Discover how managed Ray accelerates training, inference, and data processing, while Azure delivers the security, scale, and reliability needed for enterprise AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776963613993001Zbnm","RegistrantKey":"1768245294609build26-1776963613993001Zbnm"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996815001pSYL","RegistrantKey":"1768245294609build26-1774284996815001pSYL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"bf50c140-6850-4cea-9af1-572e3c727850","startDateTime":"2026-06-03T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Lachie Evenson, Subhav Mital","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493196993001nrRG","langLocale":"en-US","sessionId":"1774493196993001nrRG","sessionInstanceId":"1774493196993001nrRG","sessionCode":"TT623","sessionCodeNormalized":"EN-US-TT623","title":"Scale Python AI Faster with Anyscale on Azure","sortTitle":"scale python ai faster with anyscale on azure","description":"As Python becomes central to modern AI, many teams struggle to scale from experiments to production. Join this table talk to learn how Anyscale on Azure helps teams move from notebooks to large-scale, distributed AI workloads using open source tools—without managing infrastructure. Discover how managed Ray accelerates training, inference, and data processing, while Azure delivers the security, scale, and reliability needed for enterprise AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196993001nrRG_20260512023953744.avif","speakerIds":["1768245294609build26-1774284996815001pSYL","1768245294609build26-1776963613993001Zbnm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.641Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196993001nrRG_20260512023953744.avif","ogTitle":"Scale Python AI Faster with Anyscale on Azure","ogDescription":"As Python becomes central to modern AI, many teams struggle to scale from experiments to production. Join this table talk to learn how Anyscale on Azure helps teams move from notebooks to large-scale, distributed AI workloads using open source tools—without managing infrastructure. Discover how managed Ray accelerates training, inference, and data processing, while Azure delivers the security, scale, and reliability needed for enterprise AI.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776963613993001Zbnm","RegistrantKey":"1768245294609build26-1776963613993001Zbnm"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774284996815001pSYL","RegistrantKey":"1768245294609build26-1774284996815001pSYL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"258a6edf-256e-4b63-b9a4-630b6956c659","startDateTime":"2026-06-03T00:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maxim Lukiyanov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493196838001nfhm","langLocale":"en-US","sessionId":"1774493196838001nfhm","sessionInstanceId":"1774493196838001nfhm","sessionCode":"TT622","sessionCodeNormalized":"EN-US-TT622","title":"Scaling PostgreSQL for Agentic Apps: Patterns and Tradeoffs","sortTitle":"scaling postgresql for agentic apps: patterns and tradeoffs","description":"Agentic apps place new demands on PostgreSQL - low-latency retrieval, the need to combine graph, transactional, and document models within a single system, integration with AI models. In this Table Talk, explore architecture patterns for agentic development with PostgreSQL using Azure Horizon DB. Compare approaches for embedding pipelines, agentic architectures - balancing performance, cost, and scale - and learn where designs break in production. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"00:15 - 01:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T01:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196838001nfhm_20260512023941148.avif","speakerIds":["1768245294609build26-1765837200716001tuxr"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.642Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493196838001nfhm_20260512023941148.avif","ogTitle":"Scaling PostgreSQL for Agentic Apps: Patterns and Tradeoffs","ogDescription":"Agentic apps place new demands on PostgreSQL - low-latency retrieval, the need to combine graph, transactional, and document models within a single system, integration with AI models. In this Table Talk, explore architecture patterns for agentic development with PostgreSQL using Azure Horizon DB. Compare approaches for embedding pipelines, agentic architectures - balancing performance, cost, and scale - and learn where designs break in production. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1765837200716001tuxr","RegistrantKey":"1768245294609build26-1765837200716001tuxr"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Azure HorizonDB","logicalValue":"Azure HorizonDB"}]},{"@search.score":1.0,"documentId":"53dafdf4-7d89-467f-a6e2-c10d6d9424d1","startDateTime":"2026-06-03T21:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Carie Fisher, Aaron Gustafson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777287294165001SrQc","langLocale":"en-US","sessionId":"1777287294165001SrQc","sessionInstanceId":"1777287294165001SrQc","sessionCode":"TT690","sessionCodeNormalized":"EN-US-TT690","title":"Shifting accessibility left in AI code","sortTitle":"shifting accessibility left in ai code","description":"Explore what it means to shift accessibility left in your engineering pipeline so innovation scales inclusion by default. Learn how tools like GitHub's Accessibility Scanner bring accessibility signals directly into developer workflows alongside security and quality. We’ll open the discussion to share perspectives on what’s working, where gaps remain, and how you and your teams can embed accessibility into AI tooling and engineering practices.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"21:15 - 22:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T22:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1777287294165001SrQc_20260512031518161.avif","speakerIds":["1768245294609build26-1777300452434001pMw7","1768245294609build26-1773332637431001OOg9"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.387Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1777287294165001SrQc_20260512031518161.avif","ogTitle":"Shifting accessibility left in AI code","ogDescription":"Explore what it means to shift accessibility left in your engineering pipeline so innovation scales inclusion by default. Learn how tools like GitHub's Accessibility Scanner bring accessibility signals directly into developer workflows alongside security and quality. We’ll open the discussion to share perspectives on what’s working, where gaps remain, and how you and your teams can embed accessibility into AI tooling and engineering practices.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777300452434001pMw7","RegistrantKey":"1768245294609build26-1777300452434001pMw7"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773332637431001OOg9","RegistrantKey":"1768245294609build26-1773332637431001OOg9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"GitHub","logicalValue":"GitHub"}]},{"@search.score":1.0,"documentId":"095a1e63-2521-47ce-933d-34a2df3333d2","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Scott Kennedy","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282848001kOb4","langLocale":"en-US","sessionId":"1774493282848001kOb4","sessionInstanceId":"1774493282848001kOb4","sessionCode":"TT657","sessionCodeNormalized":"EN-US-TT657","title":"Shipping at the Speed of Ideas: Real Lessons in AI-First Development","sortTitle":"shipping at the speed of ideas: real lessons in ai-first development","description":"When the full stack goes from concept to live in hours, software development changes in ways not obvious until you're in it. We learned this firsthand building an AI software creation platform, and see it play out daily. This session shares hard lessons from both sides: engineering a system that writes reliable code, and patterns that succeed and fail when developers go AI-first. Walk away knowing where to trust the agent, where to stay in control, and how to build things to last.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282848001kOb4_20260512024232095.avif","speakerIds":["1768245294609build26-1777731730099001WAOS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.809Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282848001kOb4_20260512024232095.avif","ogTitle":"Shipping at the Speed of Ideas: Real Lessons in AI-First Development","ogDescription":"When the full stack goes from concept to live in hours, software development changes in ways not obvious until you're in it. We learned this firsthand building an AI software creation platform, and see it play out daily. This session shares hard lessons from both sides: engineering a system that writes reliable code, and patterns that succeed and fail when developers go AI-first. Walk away knowing where to trust the agent, where to stay in control, and how to build things to last.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777731730099001WAOS","RegistrantKey":"1768245294609build26-1777731730099001WAOS"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Platform","logicalValue":"Platform"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"f760bf52-23f1-41af-9243-9a2b886bc9ee","startDateTime":"2026-06-03T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Fatih Akin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778509007287001V60N","langLocale":"en-US","sessionId":"1778509007287001V60N","sessionInstanceId":"1778509007287001V60N","sessionCode":"TT600-R1","sessionCodeNormalized":"EN-US-TT600-R1","title":"Table Talk with prompts.chat creator Fatih Kadir Akin","sortTitle":"table talk with prompts.chat creator fatih kadir akin","description":"prompts.chat creator, Fatih Kadir Akin, will share more of his story creating an open-source prompt library into a broader ecosystem. He will explain why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows. Join him for this table talk and hear his perspective as an open-source leader in the age of AI. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"16:15 - 17:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778509007287001V60N_20260512032006420.avif","speakerIds":["1768245294609build26-1776361485371001HMvG"],"sessionLinks":[],"relatedSessionCodes":["LTG468"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.777Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778509007287001V60N_20260512032006420.avif","ogTitle":"Table Talk with prompts.chat creator Fatih Kadir Akin","ogDescription":"prompts.chat creator, Fatih Kadir Akin, will share more of his story creating an open-source prompt library into a broader ecosystem. He will explain why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows. Join him for this table talk and hear his perspective as an open-source leader in the age of AI. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776361485371001HMvG","RegistrantKey":"1768245294609build26-1776361485371001HMvG"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"f0f52900-df49-4e45-a886-20e8700973fd","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Fatih Akin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493194634001nej5","langLocale":"en-US","sessionId":"1774493194634001nej5","sessionInstanceId":"1774493194634001nej5","sessionCode":"TT600","sessionCodeNormalized":"EN-US-TT600","title":"Table Talk with prompts.chat creator Fatih Kadir Akin","sortTitle":"table talk with prompts.chat creator fatih kadir akin","description":"prompts.chat creator, Fatih Kadir Akin, will share more of his story creating an open-source prompt library into a broader ecosystem. He will explain why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows. Join him for this table talk and hear his perspective as an open-source leader in the age of AI. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493194634001nej5_20260512023859064.avif","speakerIds":["1768245294609build26-1776361485371001HMvG"],"sessionLinks":[],"relatedSessionCodes":["LTG468"],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.641Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493194634001nej5_20260512023859064.avif","ogTitle":"Table Talk with prompts.chat creator Fatih Kadir Akin","ogDescription":"prompts.chat creator, Fatih Kadir Akin, will share more of his story creating an open-source prompt library into a broader ecosystem. He will explain why prompts still matter in the age of agents, how they help define intent and context, and why open systems are important for collaboration, reuse, and better developer workflows. Join him for this table talk and hear his perspective as an open-source leader in the age of AI. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776361485371001HMvG","RegistrantKey":"1768245294609build26-1776361485371001HMvG"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"63e52000-ef9e-44df-a445-8369a6250b16","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dinesh Narayanan, Sai Vemprala","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282108001kcQi","langLocale":"en-US","sessionId":"1774493282108001kcQi","sessionInstanceId":"1774493282108001kcQi","sessionCode":"TT652","sessionCodeNormalized":"EN-US-TT652","title":"The Intelligence grid: Agentic AI for Real-World Robotics","sortTitle":"the intelligence grid: agentic ai for real-world robotics","description":"As robotics systems move beyond scripted autonomy toward adaptive, decision-making behavior, agentic AI is becoming the foundation for deploying intelligent robots in production environments. Join the co-founder of General Robotics and explore how they are using agentic architectures that enable robots to reason, plan, collaborate, and act across heterogeneous hardware—including robot arms, mobile robots, and drones. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282108001kcQi_20260515115335058.avif","speakerIds":["1768245294609build26-1771612404865001rzIg","1768245294609build26-1777673949408001CiJK"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.808Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282108001kcQi_20260515115335058.avif","ogTitle":"The Intelligence grid: Agentic AI for Real-World Robotics","ogDescription":"As robotics systems move beyond scripted autonomy toward adaptive, decision-making behavior, agentic AI is becoming the foundation for deploying intelligent robots in production environments. Join the co-founder of General Robotics and explore how they are using agentic architectures that enable robots to reason, plan, collaborate, and act across heterogeneous hardware—including robot arms, mobile robots, and drones. \n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"JavaScript","logicalValue":"JavaScript"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"C++","logicalValue":"C++"},{"displayValue":"C","logicalValue":"C"},{"displayValue":"TypeScript","logicalValue":"TypeScript"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777673949408001CiJK","RegistrantKey":"1768245294609build26-1777673949408001CiJK"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1771612404865001rzIg","RegistrantKey":"1768245294609build26-1771612404865001rzIg"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"}]},{"@search.score":1.0,"documentId":"6e720395-bb10-4f82-840b-a73fd5e975c0","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Brian Benz","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778524766937001Z9qa","langLocale":"en-US","sessionId":"1778524766937001Z9qa","sessionInstanceId":"1778524766937001Z9qa","sessionCode":"TT664B","sessionCodeNormalized":"EN-US-TT664B","title":"The Legacy Java Problem No One Wants to Inherit","sortTitle":"the legacy java problem no one wants to inherit","description":"Every team has one: the Java app no one fully understands, running a runtime, no one wants to touch, with tests that may or may not exist. Modernizing it used to mean months of planning and a careful rewrite. Now AI agents are quietly changing the math. This Table Talk explores how teams are tackling legacy Java modernization with agentic tools—analyzing unfamiliar codebases, planning upgrades, and refactoring at a pace that wasn’t possible a year ago. We’ll trade stories on what’s working, what’s breaking, and how the role of the developer is shifting in the process.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778524766937001Z9qa_20260512152843550.avif","speakerIds":["1768245294609build26-1776199101971001JvYe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.776Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778524766937001Z9qa_20260512152843550.avif","ogTitle":"The Legacy Java Problem No One Wants to Inherit","ogDescription":"Every team has one: the Java app no one fully understands, running a runtime, no one wants to touch, with tests that may or may not exist. Modernizing it used to mean months of planning and a careful rewrite. Now AI agents are quietly changing the math. This Table Talk explores how teams are tackling legacy Java modernization with agentic tools—analyzing unfamiliar codebases, planning upgrades, and refactoring at a pace that wasn’t possible a year ago. We’ll trade stories on what’s working, what’s breaking, and how the role of the developer is shifting in the process.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776199101971001JvYe","RegistrantKey":"1768245294609build26-1776199101971001JvYe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"bafa8cf5-3c0f-4dd4-8acc-1b2f8c75ed1c","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayan Gupta","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283881001ksjq","langLocale":"en-US","sessionId":"1774493283881001ksjq","sessionInstanceId":"1774493283881001ksjq","sessionCode":"TT664A","sessionCodeNormalized":"EN-US-TT664A","title":"The Legacy Java Problem No One Wants to Inherit","sortTitle":"the legacy java problem no one wants to inherit","description":"Every team has one: the Java app no one fully understands, running a runtime, no one wants to touch, with tests that may or may not exist. Modernizing it used to mean months of planning and a careful rewrite. Now AI agents are quietly changing the math. This Table Talk explores how teams are tackling legacy Java modernization with agentic tools—analyzing unfamiliar codebases, planning upgrades, and refactoring at a pace that wasn’t possible a year ago. We’ll trade stories on what’s working, what’s breaking, and how the role of the developer is shifting in the process.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283881001ksjq_20260512153211466.avif","speakerIds":["1768245294609build26-1767923733240001J5fY"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.839Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283881001ksjq_20260512153211466.avif","ogTitle":"The Legacy Java Problem No One Wants to Inherit","ogDescription":"Every team has one: the Java app no one fully understands, running a runtime, no one wants to touch, with tests that may or may not exist. Modernizing it used to mean months of planning and a careful rewrite. Now AI agents are quietly changing the math. This Table Talk explores how teams are tackling legacy Java modernization with agentic tools—analyzing unfamiliar codebases, planning upgrades, and refactoring at a pace that wasn’t possible a year ago. We’ll trade stories on what’s working, what’s breaking, and how the role of the developer is shifting in the process.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 3","logicalValue":"Gateway Pavilion, Level 2, Table 3"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1767923733240001J5fY","RegistrantKey":"1768245294609build26-1767923733240001J5fY"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"fe860b13-ab7a-47e6-8512-c7e269b4380d","startDateTime":"2026-06-03T17:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Egina Malaj","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493286595001kdVd","langLocale":"en-US","sessionId":"1774493286595001kdVd","sessionInstanceId":"1774493286595001kdVd","sessionCode":"TT682","sessionCodeNormalized":"EN-US-TT682","title":"Trusted AI Built for Production","sortTitle":"trusted ai built for production","description":"Responsible AI is widely discussed, but hard to operationalize. Many teams struggle to move AI systems from MVP to production. This session covers defining risk thresholds (safety, jailbreaks, accuracy, hallucinations), generating test cases with LLMs, and building golden datasets. We’ll discuss evaluating multi-turn agentic systems using Azure evaluation SDKs and AI Red Teaming Agent, plus integration into CI/CD for repeatable, production-ready systems across prompts, models, and agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"17:15 - 18:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T18:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286595001kdVd_20260512024412206.avif","speakerIds":["1768245294609build26-1776296191511001xsDz"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.84Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286595001kdVd_20260512024412206.avif","ogTitle":"Trusted AI Built for Production","ogDescription":"Responsible AI is widely discussed, but hard to operationalize. Many teams struggle to move AI systems from MVP to production. This session covers defining risk thresholds (safety, jailbreaks, accuracy, hallucinations), generating test cases with LLMs, and building golden datasets. We’ll discuss evaluating multi-turn agentic systems using Azure evaluation SDKs and AI Red Teaming Agent, plus integration into CI/CD for repeatable, production-ready systems across prompts, models, and agents.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776296191511001xsDz","RegistrantKey":"1768245294609build26-1776296191511001xsDz"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"0cb73a15-4aa7-46e5-b977-d33391869821","startDateTime":"2026-06-03T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Andrew Leader, Lei Xu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493199040001n4Wm","langLocale":"en-US","sessionId":"1774493199040001n4Wm","sessionInstanceId":"1774493199040001n4Wm","sessionCode":"TT646","sessionCodeNormalized":"EN-US-TT646","title":"Using local AI in your Windows apps: Guidance and best practices","sortTitle":"using local ai in your windows apps: guidance and best practices","description":"Join us for a table talk where we will explore the different paths for bringing AI to your apps. We'll cover pros and cons of: using local models vs. cloud and hybrid options, running models provided by catalog vs open-source, using AI APIs vs writing to models yourself, supporting web apps, and managing model deployment. This Table Talk will pair with the BRK260 breakout session to help you ship performant, scalable, and secure AI-powered Windows apps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493199040001n4Wm_20260512024107987.avif","speakerIds":["1768245294609build26-1774285002790001pcAb","1768245294609build26-1774385980999001nI1l"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.793Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493199040001n4Wm_20260512024107987.avif","ogTitle":"Using local AI in your Windows apps: Guidance and best practices","ogDescription":"Join us for a table talk where we will explore the different paths for bringing AI to your apps. We'll cover pros and cons of: using local models vs. cloud and hybrid options, running models provided by catalog vs open-source, using AI APIs vs writing to models yourself, supporting web apps, and managing model deployment. This Table Talk will pair with the BRK260 breakout session to help you ship performant, scalable, and secure AI-powered Windows apps.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774385980999001nI1l","RegistrantKey":"1768245294609build26-1774385980999001nI1l"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774285002790001pcAb","RegistrantKey":"1768245294609build26-1774285002790001pcAb"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"98b6340c-7b77-4e9c-a810-f6a448b99903","startDateTime":"2026-06-02T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Marisela Cerda, Nelly Delgado","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283291001kSMr","langLocale":"en-US","sessionId":"1774493283291001kSMr","sessionInstanceId":"1774493283291001kSMr","sessionCode":"TT660A","sessionCodeNormalized":"EN-US-TT660A","title":"What AI Can’t Replace: The Enduring Value of Human Judgment, Trust, and Connection","sortTitle":"what ai can’t replace: the enduring value of human judgment, trust, and connection","description":"As AI agents take on more planning, execution, and coordination, human contribution is shifting—not disappearing. This session explores what remains uniquely human in an agentic world, and why those capabilities will increase in value.\n\nWe’ll examine where human judgment outperforms machine optimization, how trust is built in AI-mediated systems, and why connection and accountability can’t be fully virtualized.\n\nJoin the conversation to rethink the future of work, product design, and leadership—focusing on where human value matters most.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283291001kSMr_20260520160522758.avif","speakerIds":["1768245294609build26-1770320163509001ehOc","1768245294609build26-1770330200998001uWJm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.824Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283291001kSMr_20260520160522758.avif","ogTitle":"What AI Can’t Replace: The Enduring Value of Human Judgment, Trust, and Connection","ogDescription":"As AI agents take on more planning, execution, and coordination, human contribution is shifting—not disappearing. This session explores what remains uniquely human in an agentic world, and why those capabilities will increase in value.\n\nWe’ll examine where human judgment outperforms machine optimization, how trust is built in AI-mediated systems, and why connection and accountability can’t be fully virtualized.\n\nJoin the conversation to rethink the future of work, product design, and leadership—focusing on where human value matters most.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770330200998001uWJm","RegistrantKey":"1768245294609build26-1770330200998001uWJm"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770320163509001ehOc","RegistrantKey":"1768245294609build26-1770320163509001ehOc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"658531b6-095c-4646-ba10-ecb50c59969d","startDateTime":"2026-06-02T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gerald Tiu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779217428091001lX8v","langLocale":"en-US","sessionId":"1779217428091001lX8v","sessionInstanceId":"1779217428091001lX8v","sessionCode":"TT660B","sessionCodeNormalized":"EN-US-TT660B","title":"What AI Can’t Replace: The Enduring Value of Human Judgment, Trust, and Connection","sortTitle":"what ai can’t replace: the enduring value of human judgment, trust, and connection","description":"As AI agents take on more planning, execution, and coordination, human contribution is shifting—not disappearing. This session explores what remains uniquely human in an agentic world, and why those capabilities will increase in value.\n\nWe’ll examine where human judgment outperforms machine optimization, how trust is built in AI-mediated systems, and why connection and accountability can’t be fully virtualized.\n\nJoin the conversation to rethink the future of work, product design, and leadership—focusing on where human value matters most.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779217428091001lX8v_20260520160321926.avif","speakerIds":["1768245294609build26-1751904960194001yFrX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.235Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779217428091001lX8v_20260520160321926.avif","ogTitle":"What AI Can’t Replace: The Enduring Value of Human Judgment, Trust, and Connection","ogDescription":"As AI agents take on more planning, execution, and coordination, human contribution is shifting—not disappearing. This session explores what remains uniquely human in an agentic world, and why those capabilities will increase in value.\n\nWe’ll examine where human judgment outperforms machine optimization, how trust is built in AI-mediated systems, and why connection and accountability can’t be fully virtualized.\n\nJoin the conversation to rethink the future of work, product design, and leadership—focusing on where human value matters most.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1751904960194001yFrX","RegistrantKey":"1768245294609build26-1751904960194001yFrX"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"23424485-4147-4929-8cb2-bdf9d9ce657b","startDateTime":"2026-06-03T16:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Bennett Gould, Grant Griffiths","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493282406001kRzC","langLocale":"en-US","sessionId":"1774493282406001kRzC","sessionInstanceId":"1774493282406001kRzC","sessionCode":"TT654","sessionCodeNormalized":"EN-US-TT654","title":"Why is debugging distributed, complex production systems still this hard?","sortTitle":"why is debugging distributed, complex production systems still this hard?","description":"Azure native applications built on AKS produce logs, metrics, and traces across highly distributed systems. During an incident, correlating that data across services, dependencies, and recent changes becomes the hardest part of the problem. This discussion focuses on how developers approach that challenge today and how it can be improved.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"16:15 - 17:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T17:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282406001kRzC_20260519172819937.avif","speakerIds":["1768245294609build26-1777930702053001Vurm","1768245294609build26-1773644062745001YiSc"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.809Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493282406001kRzC_20260519172819937.avif","ogTitle":"Why is debugging distributed, complex production systems still this hard?","ogDescription":"Azure native applications built on AKS produce logs, metrics, and traces across highly distributed systems. During an incident, correlating that data across services, dependencies, and recent changes becomes the hardest part of the problem. This discussion focuses on how developers approach that challenge today and how it can be improved.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1773644062745001YiSc","RegistrantKey":"1768245294609build26-1773644062745001YiSc"},{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777930702053001Vurm","RegistrantKey":"1768245294609build26-1777930702053001Vurm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"AKS","logicalValue":"AKS"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"b077b9aa-e88a-476e-bb8a-457e29811796","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Marisela Cerda","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493283435001kybx","langLocale":"en-US","sessionId":"1774493283435001kybx","sessionInstanceId":"1774493283435001kybx","sessionCode":"TT661A","sessionCodeNormalized":"EN-US-TT661A","title":"Women in AI: Connecting on Innovation, Change, and Impact","sortTitle":"women in ai: connecting on innovation, change, and impact","description":"This table talk creates an opportunity for women to connect and share what’s on your mind as AI continues to reshape the way we work. The conversation will explore perspectives on AI development, tooling, and innovation. It will also examine the broader change and transformation driven by AI and where women see the greatest opportunities to influence its future.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283435001kybx_20260514205053894.avif","speakerIds":["1768245294609build26-1770320163509001ehOc"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.823Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493283435001kybx_20260514205053894.avif","ogTitle":"Women in AI: Connecting on Innovation, Change, and Impact","ogDescription":"This table talk creates an opportunity for women to connect and share what’s on your mind as AI continues to reshape the way we work. The conversation will explore perspectives on AI development, tooling, and innovation. It will also examine the broader change and transformation driven by AI and where women see the greatest opportunities to influence its future.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 4","logicalValue":"Gateway Pavilion, Level 2, Table 4"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770320163509001ehOc","RegistrantKey":"1768245294609build26-1770320163509001ehOc"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"9e546feb-3165-4f3f-92d9-2cfbea6f757b","startDateTime":"2026-06-02T22:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nelly Delgado","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778277467022001vE7w","langLocale":"en-US","sessionId":"1778277467022001vE7w","sessionInstanceId":"1778277467022001vE7w","sessionCode":"TT661B","sessionCodeNormalized":"EN-US-TT661B","title":"Women in AI: Connecting on Innovation, Change, and Impact","sortTitle":"women in ai: connecting on innovation, change, and impact","description":"This table talk creates an opportunity for women to connect and share what’s on your mind as AI continues to reshape the way we work. The conversation will explore perspectives on AI development, tooling, and innovation. It will also examine the broader change and transformation driven by AI and where women see the greatest opportunities to influence its future.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"22:15 - 23:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T23:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778277467022001vE7w_20260514205403557.avif","speakerIds":["1768245294609build26-1770330200998001uWJm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.776Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778277467022001vE7w_20260514205403557.avif","ogTitle":"Women in AI: Connecting on Innovation, Change, and Impact","ogDescription":"This table talk creates an opportunity for women to connect and share what’s on your mind as AI continues to reshape the way we work. The conversation will explore perspectives on AI development, tooling, and innovation. It will also examine the broader change and transformation driven by AI and where women see the greatest opportunities to influence its future.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1770330200998001uWJm","RegistrantKey":"1768245294609build26-1770330200998001uWJm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"6beb3ad1-1f03-4fde-bfba-eb6911ecf5a3","startDateTime":"2026-06-02T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Trevor Back","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493197483001nJ6W","langLocale":"en-US","sessionId":"1774493197483001nJ6W","sessionInstanceId":"1774493197483001nJ6W","sessionCode":"TT630","sessionCodeNormalized":"EN-US-TT630","title":"Working with Models - from prompt to production","sortTitle":"working with models - from prompt to production","description":"How do you turn AI model capability into real product value?\u00A0We’ll cover the essentials of working with modern AI models — from choosing the right model and shaping outputs, to grounding with data and running at scale. We’ll discuss practical patterns for prompting, RAG, evaluation, and orchestration, plus the realities of cost, safety, and reliability. Bring your use cases—this will be a hands-on, discussion-driven session focused on what actually works in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197483001nJ6W_20260513131330471.avif","speakerIds":["1768245294609build26-1777926803337001VfEp"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.642Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493197483001nJ6W_20260513131330471.avif","ogTitle":"Working with Models - from prompt to production","ogDescription":"How do you turn AI model capability into real product value?\u00A0We’ll cover the essentials of working with modern AI models — from choosing the right model and shaping outputs, to grounding with data and running at scale. We’ll discuss practical patterns for prompting, RAG, evaluation, and orchestration, plus the realities of cost, safety, and reliability. Bring your use cases—this will be a hands-on, discussion-driven session focused on what actually works in production.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 2","logicalValue":"Gateway Pavilion, Level 2, Table 2"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1777926803337001VfEp","RegistrantKey":"1768245294609build26-1777926803337001VfEp"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"8ad8bf31-fdaa-41b0-86a6-cd429f31a370","startDateTime":"2026-06-02T20:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vladimir Gusarov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774493286453001kVu5","langLocale":"en-US","sessionId":"1774493286453001kVu5","sessionInstanceId":"1774493286453001kVu5","sessionCode":"TT681","sessionCodeNormalized":"EN-US-TT681","title":"Your Backlog Is Lying: Enforce It with PR Checks","sortTitle":"your backlog is lying: enforce it with pr checks","description":"If your backlog isn’t consistent, your metrics and plans aren’t either. In this session, we’ll explore how to enforce backlog integrity using Pull Request checks in Azure Repos and GitHub. See how to validate work items, enforce structure, and block merges when standards aren’t met—embedding governance directly into your SDLC without slowing developers down.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"20:15 - 21:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-02T21:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286453001kVu5_20260512024359200.avif","speakerIds":["1768245294609build26-1774633213234001AzOm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.841Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774493286453001kVu5_20260512024359200.avif","ogTitle":"Your Backlog Is Lying: Enforce It with PR Checks","ogDescription":"If your backlog isn’t consistent, your metrics and plans aren’t either. In this session, we’ll explore how to enforce backlog integrity using Pull Request checks in Azure Repos and GitHub. See how to validate work items, enforce structure, and block merges when standards aren’t met—embedding governance directly into your SDLC without slowing developers down.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 5","logicalValue":"Gateway Pavilion, Level 2, Table 5"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774633213234001AzOm","RegistrantKey":"1768245294609build26-1774633213234001AzOm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"528e90ad-fdbe-4475-9626-0fe1d002e5fa","startDateTime":"2026-06-02T23:15:00Z","partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Vladimir Gusarov","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779812499556001rojs","langLocale":"en-US","sessionId":"1779812499556001rojs","sessionInstanceId":"1779812499556001rojs","sessionCode":"TT681-R1","sessionCodeNormalized":"EN-US-TT681-R1","title":"Your Backlog Is Lying: Enforce It with PR Checks","sortTitle":"your backlog is lying: enforce it with pr checks","description":"If your backlog isn’t consistent, your metrics and plans aren’t either. In this session, we’ll explore how to enforce backlog integrity using Pull Request checks in Azure Repos and GitHub. See how to validate work items, enforce structure, and block merges when standards aren’t met—embedding governance directly into your SDLC without slowing developers down.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":7,"TimeSlot":"23:15 - 00:00","startDateTimeNullGroup":1,"endDateTime":"2026-06-03T00:00:00Z","durationInMinutes":45,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812499556001rojs_20260527112601409.avif","speakerIds":["1768245294609build26-1774633213234001AzOm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.237Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1779812499556001rojs_20260527112601409.avif","ogTitle":"Your Backlog Is Lying: Enforce It with PR Checks","ogDescription":"If your backlog isn’t consistent, your metrics and plans aren’t either. In this session, we’ll explore how to enforce backlog integrity using Pull Request checks in Azure Repos and GitHub. See how to validate work items, enforce structure, and block merges when standards aren’t met—embedding governance directly into your SDLC without slowing developers down.\n\nSeating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.","deliveryTypes":[{"displayValue":"In-person","logicalValue":"In-person"}],"sessionType":{"displayValue":"Table Talk","logicalValue":"Table Talk"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"Gateway Pavilion, Level 2, Table 1","logicalValue":"Gateway Pavilion, Level 2, Table 1"},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774633213234001AzOm","RegistrantKey":"1768245294609build26-1774633213234001AzOm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Community","logicalValue":"Community"}]},{"@search.score":1.0,"documentId":"8266579c-7db9-4a60-a0cb-bc0c2f90126b","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778247732591001qg9J","langLocale":"en-US","sessionId":"1778247732591001qg9J","sessionInstanceId":"1778247732591001qg9J","sessionCode":"LAB530D","sessionCodeNormalized":"EN-US-LAB530D","title":"Digital Lab: Engineering agents that reason, act, and adapt","sortTitle":"digital lab: engineering agents that reason, act, and adapt","description":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778247732591001qg9J_20260512033028627.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.773Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778247732591001qg9J_20260512033028627.avif","ogTitle":"Digital Lab: Engineering agents that reason, act, and adapt","ogDescription":"Lost in San Francisco with a deadline to make it to Build, you’ll code an AI agent using Foundry Agent Service, Foundry IQ, and Foundry Tools. Step by step, you’ll design an agent that reasons, acts, and adapts in real time—integrating MCP tools to solve challenges along the way. Leave with hands‑on experience building and orchestrating production‑ready agentic systems.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"2f1fb46e-3f33-456b-af12-6ebbafad5949","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778250717137001I9sZ","langLocale":"en-US","sessionId":"1778250717137001I9sZ","sessionInstanceId":"1778250717137001I9sZ","sessionCode":"LAB501D","sessionCodeNormalized":"EN-US-LAB501D","title":"Digital Lab: From zero to deployed on Azure with AI agents","sortTitle":"digital lab: from zero to deployed on azure with ai agents","description":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778250717137001I9sZ_20260512033208262.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":["BRK207"],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.773Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778250717137001I9sZ_20260512033208262.avif","ogTitle":"Digital Lab: From zero to deployed on Azure with AI agents","ogDescription":"What happens when you let AI agents do the building? In this hands-on lab, you'll go from an empty terminal to a deployed app on Azure — with GitHub Copilot CLI and coding agents handling the scaffolding, coding, debugging, and deployment. You'll use the new Azure skills to provision resources and wire up services through natural language, no portal required. This isn't a demo you watch. You'll walk out with a real, working dev workflow you can take straight to your next project.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"239f0a80-bd09-40ce-85a0-ef02b4d921bc","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778252978261001PCVD","langLocale":"en-US","sessionId":"1778252978261001PCVD","sessionInstanceId":"1778252978261001PCVD","sessionCode":"LAB520D","sessionCodeNormalized":"EN-US-LAB520D","title":"Digital Lab: Get Started with Models in Microsoft Foundry to Build AI Apps","sortTitle":"digital lab: get started with models in microsoft foundry to build ai apps","description":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778252978261001PCVD_20260512153601017.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":["BRK230"],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.775Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778252978261001PCVD_20260512153601017.avif","ogTitle":"Digital Lab: Get Started with Models in Microsoft Foundry to Build AI Apps","ogDescription":"In this hands-on lab, you will build a production-ready AI application using Microsoft Foundry, with no fine-tuning or deep machine learning expertise required. You will discover and select models, provision a Foundry project, and connect to a hosted model using the OpenAI SDK. You’ll implement a comment moderation workflow, compare model outputs, and package the solution as a hosted agent using Python, ready for real-world integration.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"c610fb33-9a31-4118-86f3-8ea9af43ba67","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778253073732001PLLA","langLocale":"en-US","sessionId":"1778253073732001PLLA","sessionInstanceId":"1778253073732001PLLA","sessionCode":"LAB521D","sessionCodeNormalized":"EN-US-LAB521D","title":"Digital Lab: Improving agent behavior using reinforcement learning from traces","sortTitle":"digital lab: improving agent behavior using reinforcement learning from traces","description":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778253073732001PLLA_20260512153614226.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":["BRK231"],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.774Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778253073732001PLLA_20260512153614226.avif","ogTitle":"Digital Lab: Improving agent behavior using reinforcement learning from traces","ogDescription":"AI agents often work, but their behavior can be inconsistent or hard to control with prompts alone. In this hands-on lab, you will start with a working agent, inspect execution traces, define “good behavior” using evaluation graders, and apply reinforcement learning to reinforce better decisions. By the end, you’ll have a practical workflow for improving agent behavior using feedback from agent runs, without redesigning your agent from scratch.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"6a78b177-53cf-4127-9094-617b9cd2a8f3","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Nitya Narasimhan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778252419747001Kkwg","langLocale":"en-US","sessionId":"1778252419747001Kkwg","sessionInstanceId":"1778252419747001Kkwg","sessionCode":"LAB540D","sessionCodeNormalized":"EN-US-LAB540D","title":"Digital Lab: Observe, optimize and protect your hosted agents in Microsoft Foundry","sortTitle":"digital lab: observe, optimize and protect your hosted agents in microsoft foundry","description":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778252419747001Kkwg_20260512033235291.avif","speakerIds":["1768245294609build26-1756141269010001LleM"],"sessionLinks":[],"relatedSessionCodes":["BRK252"],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.774Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778252419747001Kkwg_20260512033235291.avif","ogTitle":"Digital Lab: Observe, optimize and protect your hosted agents in Microsoft Foundry","ogDescription":"Modern agents fail in ways traditional monitoring can’t catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites (auto-generated evaluators + test datasets) wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Content Lead, Session Owner","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Evaluations","logicalValue":"Evaluations"}]},{"@search.score":1.0,"documentId":"bebfface-3b46-425e-ab55-664d0501726e","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778250652997001ZFEq","langLocale":"en-US","sessionId":"1778250652997001ZFEq","sessionInstanceId":"1778250652997001ZFEq","sessionCode":"LAB500D","sessionCodeNormalized":"EN-US-LAB500D","title":"Digital Lab: Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","sortTitle":"digital lab: terminal to party: live-coding with github copilot cli and hydra","description":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":8,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1778250652997001ZFEq_20260512033152374.avif","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":["BRK203"],"hasTeamsMeeting":true,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.774Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1778250652997001ZFEq_20260512033152374.avif","ogTitle":"Digital Lab: Terminal to Party: Live-Coding with GitHub Copilot CLI and Hydra","ogDescription":"Experience GitHub Copilot CLI while exploring OSS Hydra, a live-coding audio/visual environment that runs directly in the browser. Learn how to use agentic loops to create an AV party of your very own. EDM? Lo-fi? Have it all! You'll find your groove while exploring all that Copilot CLI has to offer, from planning, to managing multiple agents, to delegating longer tasks to the cloud. What better way to learn - and discover your new dev soundtrack! Warning. This program may contain flashing lights, strobe lights, loud noise, and/or smoke machines which may not be suitable for those with sensitivities.\n\nThis is a self-paced, digitally-delivered lab experience that you complete on your own schedule. Click the 'Launch' button to start the lab, and then after authentication, follow the instructions listed on the lab page to complete the activity. For technical issues or support, please contact the support team via this link: https://skill.info/build26labsupport.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Digital Lab","logicalValue":"Digital Lab"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will not be recorded","logicalValue":"Will not be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"GitHub Actions","logicalValue":"GitHub Actions"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"764672f4-6dc4-4399-9c2c-17092658c41c","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780419659229001P5in","langLocale":"en-US","sessionId":"1780419659229001P5in","sessionInstanceId":"1780419659229001P5in","sessionCode":"OD806","sessionCodeNormalized":"EN-US-OD806","title":".NET 11 in depth: Runtime, libraries, and SDK for the AI era","sortTitle":".net 11 in depth: runtime, libraries, and sdk for the ai era","description":".NET 11 delivers a new wave of improvements across the runtime, libraries, and SDK to help developers build modern applications for the AI era. In this session, we’ll take an in-depth look at the key investments in performance, diagnostics, and developer productivity, and how they come together to support intelligent, cloud-connected, and agent-driven apps.","aiDescription":"<p><strong>Introduction and Agenda:</strong> The session opens at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a> with hosts Chet Hosk and Rich Lander welcoming viewers to Microsoft Build 2026, presenting an overview of .NET 11 runtime, libraries, and SDK improvements. Chet outlines that the presentation will cover SDK and tooling-level updates followed by a deep dive into runtime changes from Rich. The team commits to showcasing enhancements coming across performance, usability, and acquisition of .NET tooling for developers.</p>\n\n<p><strong>SDK and Tooling Enhancements:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:00:51\">00:00:51</a>, Chet explains three major .NET SDK workstreams for the .NET 11 release—new user and CLI capabilities, performance optimization, and streamlined management of developer tools. Key improvements include CLI feature updates for multitargeted projects and faster command execution in agent and multi-agent environments. Major updates in this section revolve around Maui device-specific capabilities, where <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:19\">00:03:19</a> new .NET run enhancements now let developers deploy directly to target devices and emulators, extending to frameworks like Uno or Avalonia. Chet also shares efforts to make the CLI aware of AI or automated agent contexts to optimize outputs for low-latency scenarios, improving efficiency for builds and concurrent tasks. Additional focus areas include multi-threaded MSBuild, conversion of CLI tools to native AOT for reduced latency, and ecosystem-wide efforts for trimming compatibility in .NET libraries (<a class=\"medius-aisummary\" data-starttime=\"00:06:08\">00:06:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:47\">00:09:47</a>).</p>\n\n<p><strong>Performance, Acquisition, and Packaging Improvements:</strong> Between <a class=\"medius-aisummary\" data-starttime=\"00:11:06\">00:11:06</a> and <a class=\"medius-aisummary\" data-starttime=\"00:14:38\">00:14:38</a>, Chet highlights the ongoing conversion of .NET CLI commands into native AOT executables for better responsiveness and startup time. This work not only boosts performance but helps ensure ecosystem libraries meet trim-friendly requirements. The team also unveils `.NET Up`, a native AOT acquisition tool that simplifies SDK version management across platforms without administrative permissions. Additionally, improvements to SDK packaging processes through the use of file hard-links have cut SDK container sizes by about 80MB, improving download and deployment speeds across environments. These packaging optimizations build upon earlier runtime features and provide a smaller, more efficient SDK footprint for developers working in containerized setups.</p>\n\n<p><strong>Runtime and Library Innovations:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>, Rich Lander takes over to discuss runtime and library enhancements. He starts with updates to the Process API, offering new convenience methods like RunAndCaptureTextAsync and ReadAllLinesAsync to avoid common deadlocks and simplify pipe handling between processes. Rich demonstrates how the new CreateAnonymousPipe API modernizes inter-process communication in C#, alongside Fire-and-Forget execution patterns and optimized trimming behavior. From <a class=\"medius-aisummary\" data-starttime=\"00:20:13\">00:20:13</a> onward, text processing improvements center on Unicode and Regex capabilities, such as UTF-8/16 validation and universal line-ending support, ensuring developers can process text consistently across platforms. Updates also include Rune-aware string functions to correctly handle extended Unicode characters, improving emojis and complex scripts handling across applications.</p>\n\n<p><strong>System.Text.Json and Compression Advances:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:24:11\">00:24:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:29\">00:28:29</a>, Rich describes new System.Text.Json serializer policies allowing more granular control over property casing, omission of null values, and per-type serialization rules. These enhancements ease working with structured data in event-logging environments and agent-based applications. The team introduces JSON Lines (NDJSON) support to efficiently handle streaming data or AI agent conversations. Compression libraries also gain integrated Zstandard algorithms and broad optimizations for multiple codecs, delivering smaller, faster compression results that developers can measure using new benchmarks within the runtime repository.</p>\n\n<p><strong>Runtime Async and Memory Safety Projects:</strong> The final section (<a class=\"medius-aisummary\" data-starttime=\"00:28:33\">00:28:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:54\">00:43:54</a>) turns to major runtime architecture changes. Rich introduces \"Runtime Async,\" a new performance-oriented model for async calls that eliminates traditional compiler state machines. Enabled via a simple configuration property, Runtime Async delivers cleaner stack traces, smaller binaries, and potentially faster code without any language syntax changes. The team also previews the memory safety initiative beginning in .NET 11 and continuing into .NET 12, redesigning unsafe code usage into reviewable contracts. Enhancements to JIT optimizations improve bounds-check eliminations and inlining behavior, removing unsafe patterns from performance-critical libraries like SIMD. The talk closes with Rich and Chet emphasizing how developers can experiment with these preview features now, provide GitHub feedback, and help optimize .NET 11 for its upcoming general release.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.464Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/06f91e84-cd18-4043-b4b6-c4ac9b0d2b0e","ogTitle":".NET 11 in depth: Runtime, libraries, and SDK for the AI era","ogDescription":".NET 11 delivers a new wave of improvements across the runtime, libraries, and SDK to help developers build modern applications for the AI era. In this session, we’ll take an in-depth look at the key investments in performance, diagnostics, and developer productivity, and how they come together to support intelligent, cloud-connected, and agent-driven apps.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[{"link":"https://get.dot.net/11","backpackLink":null,"label":"CTA","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"c6f6468c-e338-4ca8-912b-dbd0a67968ca","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roop Kiran Chevuri, Phil Gerity","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774550685532001Q145","langLocale":"en-US","sessionId":"1774550685532001Q145","sessionInstanceId":"1774550685532001Q145","sessionCode":"OD855","sessionCodeNormalized":"EN-US-OD855","title":"Accelerating developer productivity with Windows 365","sortTitle":"accelerating developer productivity with windows 365","description":"This session explores how Windows 365 delivers secure, automated cloud experience for development. It covers developer ready images, performance enhancements, and professional grade Cloud PC capabilities. Learn what’s available today and how to use Windows 365 Cloud PCs as a flexible development platform across devices and locations—without sacrificing enterprise security.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9458a91b-4d66-456f-b4db-4f278c414e1c","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9458a91b-4d66-456f-b4db-4f278c414e1c","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9458a91b-4d66-456f-b4db-4f278c414e1c","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9458a91b-4d66-456f-b4db-4f278c414e1c","speakerIds":["1768245294609build26-1768591865058001hGD2","1768245294609build26-1774886382997001gmhe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.842Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9458a91b-4d66-456f-b4db-4f278c414e1c","ogTitle":"Accelerating developer productivity with Windows 365","ogDescription":"This session explores how Windows 365 delivers secure, automated cloud experience for development. It covers developer ready images, performance enhancements, and professional grade Cloud PC capabilities. Learn what’s available today and how to use Windows 365 Cloud PCs as a flexible development platform across devices and locations—without sacrificing enterprise security.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1768591865058001hGD2","RegistrantKey":"1768245294609build26-1768591865058001hGD2"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1774886382997001gmhe","RegistrantKey":"1768245294609build26-1774886382997001gmhe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"Azure Local","logicalValue":"Azure Local"}]},{"@search.score":1.0,"documentId":"135425d1-64d0-4593-8c0c-1984a78090a7","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Emily Lisa, Sujata Narayana","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693207751001FlUz","langLocale":"en-US","sessionId":"1775693207751001FlUz","sessionInstanceId":"1775693207751001FlUz","sessionCode":"OD817","sessionCodeNormalized":"EN-US-OD817","title":"Agentic analytics with Power BI and Microsoft Fabric","sortTitle":"agentic analytics with power bi and microsoft fabric","description":"AI-powered agents are reshaping how we build analytics solutions. Agentic analytics not only helps accelerate development cycles but also helps make it easier to apply best practices and standards across your analytics projects. In this session you'll see the latest innovation when it comes to agentic development with Power BI and Microsoft Fabric. You'll also see educational demos how to best leverage AI agents to build end-to-end enterprise-grade data experiences.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4d5ab49e-06c9-4bbc-824e-4952ba249379","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4d5ab49e-06c9-4bbc-824e-4952ba249379","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4d5ab49e-06c9-4bbc-824e-4952ba249379","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4d5ab49e-06c9-4bbc-824e-4952ba249379","speakerIds":["1768245294609build26-1776284920806001LbFB","1768245294609build26-1776284982056001K8Wy"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4d5ab49e-06c9-4bbc-824e-4952ba249379","ogTitle":"Agentic analytics with Power BI and Microsoft Fabric","ogDescription":"AI-powered agents are reshaping how we build analytics solutions. Agentic analytics not only helps accelerate development cycles but also helps make it easier to apply best practices and standards across your analytics projects. In this session you'll see the latest innovation when it comes to agentic development with Power BI and Microsoft Fabric. You'll also see educational demos how to best leverage AI agents to build end-to-end enterprise-grade data experiences.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284920806001LbFB","RegistrantKey":"1768245294609build26-1776284920806001LbFB"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284982056001K8Wy","RegistrantKey":"1768245294609build26-1776284982056001K8Wy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"8e8c23ab-db44-434b-9423-067eb1c9f1fc","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776704674660001mxnF","langLocale":"en-US","sessionId":"1776704674660001mxnF","sessionInstanceId":"1776704674660001mxnF","sessionCode":"ODSP933","sessionCodeNormalized":"EN-US-ODSP933","title":"Agentic infrastructure needs agentic observability","sortTitle":"agentic infrastructure needs agentic observability","description":"Observability pipelines were built for a world where engineers manually standardized logs, traces, and metrics. That model is breaking down as AI generates services and infrastructure faster than pipelines can keep up. Explore a new model where observability layers reason over telemetry and adapt as systems evolve. Agent-driven workflows continuously interpret and normalize data, raising new questions around guardrails, validation, and autonomy.","aiDescription":"<p><strong>Introduction – The End of Human-Centric Observability:</strong> At the beginning of the talk <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:49\">00:00:49</a>, Jimmy Herbert introduces a fundamental shift in the world of software operations. He recalls how observability tools were historically designed for human operators to debug and understand systems through logs, traces, dashboards, and alerts. That model, he argues, is breaking down because the consumers of observability are no longer people but intelligent agents capable of writing and executing code autonomously. As he declares, “agentic infrastructure needs agentic observability,” marking the thematic foundation for the rest of his presentation. He clarifies that this is not a product pitch but a systemic diagnosis of what’s failing in traditional observability since AI entered the software lifecycle <a class=\"medius-aisummary\" data-starttime=\"00:00:49\">00:00:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:20\">00:01:20</a>.</p>\n\n<p><strong>Problem One – Logs and Understanding Agent Reasoning:</strong> Herbert begins diagnosing the first major issue—logging. Traditional logs were designed for discrete, human-readable events like \"request completed\" or \"database failed\" <a class=\"medius-aisummary\" data-starttime=\"00:01:30\">00:01:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>. But intelligent agents generate something different: decision breadcrumbs, or as he terms it, \"reasoning trails.\" These trails involve document retrievals, relevance scoring, retries, and contextual reevaluation, all across long time horizons <a class=\"medius-aisummary\" data-starttime=\"00:01:55\">00:01:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:31\">00:02:31</a>. The challenge shifts from asking “what happened?” to “why did the agent decide this?” Current observability platforms aren’t designed to answer such questions because they assume deterministic systems with reconstructable narratives. In reality, similar inputs now yield different outcomes, and understanding the reasoning path has become more critical than tracking technical execution <a class=\"medius-aisummary\" data-starttime=\"00:02:45\">00:02:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:10\">00:03:10</a>.</p>\n\n<p><strong>Problem Two – Traces, Sampling, and the End of 200 OK:</strong> The second major issue Herbert identifies concerns traces and sampling strategies. He explains that traditional application performance monitoring (APM) assumes a deterministic world; however, “sampling breaks and status codes lie” <a class=\"medius-aisummary\" data-starttime=\"00:03:14\">00:03:14</a>. A random trace from an agentic workflow provides little insight because crucial reasoning steps may be excluded. Tail-based sampling also fails, since retaining thousands of spans in memory is computationally unscalable <a class=\"medius-aisummary\" data-starttime=\"00:03:37\">00:03:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:54\">00:03:54</a>. Herbert provocatively points out that a “200 OK means nothing” in this new world; it only confirms that nothing crashed, not that the outcome was correct <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:20\">00:04:20</a>. He describes how agents can hallucinate or misinterpret data despite producing successful infrastructure responses. Moreover, the explosion of telemetry from agentic systems threatens cost models and forces undesirable trade-offs between complete visibility and affordability <a class=\"medius-aisummary\" data-starttime=\"00:04:44\">00:04:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:18\">00:05:18</a>.</p>\n\n<p><strong>Problem Three – Sensitivity, Instrumentation, and Human Limits:</strong> Moving into the third challenge <a class=\"medius-aisummary\" data-starttime=\"00:05:27\">00:05:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:11\">00:06:11</a>, Herbert dives into the growing sensitivity of telemetry. Modern AI workloads embed prompts, personal data, and proprietary business logic within traces and logs. What once were technical artifacts now contain deeply private information, transforming telemetry into a liability rather than a cost item. Simultaneously, the practice of manual instrumentation collapses under the pace of AI-generated artifacts <a class=\"medius-aisummary\" data-starttime=\"00:06:08\">00:06:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:22\">00:06:22</a>. He argues for observability that instruments itself automatically and can keep pace with systems evolving faster than humans. This leads to a stark realization: traditional observability engineering has been optimized around human cognitive limits—reducing data complexity, simplifying dashboards, and filtering noise—but these optimizations now constrain AI agents that thrive on more data, more context, and more completeness <a class=\"medius-aisummary\" data-starttime=\"00:06:42\">00:06:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:11\">00:07:11</a>.</p>\n\n<p><strong>Agentic Architecture – Rebuilding Observability:</strong> Herbert then discusses how observability must evolve into an architecture designed for data abundance rather than scarcity <a class=\"medius-aisummary\" data-starttime=\"00:07:16\">00:07:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a>. He introduces Groundcover’s approach, emphasizing three principles: BYOC (Bring Your Own Cloud) to keep sensitive data under customer control, zero instrumentation for low operational friction, and unified visibility for both human- and AI-written code <a class=\"medius-aisummary\" data-starttime=\"00:08:47\">00:08:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:46\">00:09:46</a>. Central to this architecture is a separation between the model and the analytic engine. The LLM should handle understanding intent and orchestration, while deterministic back-end analytics handle computation <a class=\"medius-aisummary\" data-starttime=\"00:09:50\">00:09:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:30\">00:10:30</a>. This division allows investigation to move from manual navigation to AI-driven goal-based diagnosis—agents can autonomously find the problem, explain what went wrong, and even initiate corrective actions <a class=\"medius-aisummary\" data-starttime=\"00:10:47\">00:10:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:40\">00:11:40</a>.</p>\n\n<p><strong>Conclusion – Observability as the Agentic Operating System:</strong> In his closing remarks <a class=\"medius-aisummary\" data-starttime=\"00:12:06\">00:12:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>, Herbert raises governance concerns: how much autonomy should these observability agents have, who approves their interventions, and who holds accountability for automated fixes gone wrong? Despite the uncertainties, he predicts that companies mastering “agentic observability” will hold a structural advantage. The overarching principle—“AI cannot fix what it cannot see”—encapsulates his thesis. Observability, he concludes, is transforming into the operating system for the agentic software development lifecycle (SDLC). It will become not merely a passive monitoring layer but an active, intelligent participant in understanding and maintaining modern software systems—a necessary foundation for any AI-driven infrastructure to function effectively.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/6f489fc7-7dde-4a5a-b907-69f3f69aee99","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/6f489fc7-7dde-4a5a-b907-69f3f69aee99","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/6f489fc7-7dde-4a5a-b907-69f3f69aee99","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/6f489fc7-7dde-4a5a-b907-69f3f69aee99","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.383Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["groundcover"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/6f489fc7-7dde-4a5a-b907-69f3f69aee99","ogTitle":"Agentic infrastructure needs agentic observability","ogDescription":"Observability pipelines were built for a world where engineers manually standardized logs, traces, and metrics. That model is breaking down as AI generates services and infrastructure faster than pipelines can keep up. Explore a new model where observability layers reason over telemetry and adapt as systems evolve. Agent-driven workflows continuously interpret and normalize data, raising new questions around guardrails, validation, and autonomy.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"}]},{"@search.score":1.0,"documentId":"04d1853e-cec5-4147-84f2-8fd92ca558eb","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780419294588001UqMp","langLocale":"en-US","sessionId":"1780419294588001UqMp","sessionInstanceId":"1780419294588001UqMp","sessionCode":"OD805","sessionCodeNormalized":"EN-US-OD805","title":"AI Building Blocks for .NET: Add intelligence to your C# apps","sortTitle":"ai building blocks for .net: add intelligence to your c# apps","description":"A practical, opinionated guide to building intelligent apps in .NET.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video opens with Bruno welcoming viewers and introducing the session on AI in .NET <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>. He explains that the goal is to explore AI building blocks provided by Microsoft for developers and demonstrates their integration using a showcase application, a full support center app that analyzes incidents and user feedback <a class=\"medius-aisummary\" data-starttime=\"00:00:31\">00:00:31</a>. Bruno mentions that the application runs with Aspire coordination, showing how different services interact including a Blazor web UI, a Microsoft Agent Framework agent for image generation, and an NVIDIA Nemo agent for data analysis <a class=\"medius-aisummary\" data-starttime=\"00:02:11\">00:02:11</a>. He introduces Aspire traces to monitor service interactions and sets the context for the upcoming deep dive into the .NET AI architecture.</p>\n\n<p><strong>Microsoft Extensions for AI and Local Deployment:</strong> Bruno then introduces Microsoft Extensions for AI, explaining its role as a chat abstraction layer that allows easy interaction with large language models (LLMs) via standardized interfaces <a class=\"medius-aisummary\" data-starttime=\"00:03:41\">00:03:41</a>. He showcases sample applications using Azure Foundry with integrated security, avoiding the need for API keys while maintaining authentication with CLI credentials <a class=\"medius-aisummary\" data-starttime=\"00:06:22\">00:06:22</a>. A live demo illustrates switching between models—like GPT-5 and Kimi—and retrieving model-specific responses. Bruno moves on to demonstrate running local models using Ollama <a class=\"medius-aisummary\" data-starttime=\"00:08:10\">00:08:10</a>, performing sentiment analysis locally through the same interface. He emphasizes how Microsoft’s chat client can connect seamlessly to either cloud or local models, maintaining consistent workflows across environments <a class=\"medius-aisummary\" data-starttime=\"00:10:09\">00:10:09</a>.</p>\n\n<p><strong>Data and Embeddings Fundamentals:</strong> Moving deeper, Bruno explains using data within AI applications, particularly Retrieval-Augmented Generation (RAG), which involves reading, chunking, embedding, and storing text <a class=\"medius-aisummary\" data-starttime=\"00:11:01\">00:11:01</a>. He demonstrates how embeddings convert text into vector representations for semantic search and shows an example where movie descriptions are embedded in memory for querying <a class=\"medius-aisummary\" data-starttime=\"00:14:09\">00:14:09</a>. By asking questions like “find a family-friendly movie with dragons,” the app accurately retrieves related content such as “Shrek” or “The Matrix.” He also presents a Markdown data ingestion pipeline <a class=\"medius-aisummary\" data-starttime=\"00:20:04\">00:20:04</a>, detailing how documents can be read, chunked, summarized, and written into a vector store using SQLite. This section demonstrates the scalable flexibility of working with embeddings locally or in the cloud, bridging unstructured data to semantic search capabilities through .NET AI libraries.</p>\n\n<p><strong>Integrating External Tools with MCP:</strong> Bruno transitions to introducing the Model Context Protocol (MCP), a standard enabling AI applications to access external tools and APIs <a class=\"medius-aisummary\" data-starttime=\"00:23:01\">00:23:01</a>. He demonstrates a console app connecting to the Microsoft Learn MCP server, showing how chat clients invoke MCP tools to retrieve live documentation and current framework versions from learn.microsoft.com <a class=\"medius-aisummary\" data-starttime=\"00:25:02\">00:25:02</a>. A side-by-side test shows how answers improve when connected via MCP, bringing live URLs and precise version details without manual linking <a class=\"medius-aisummary\" data-starttime=\"00:29:04\">00:29:04</a>. This integration showcases how MCP empowers AI clients to extend their capabilities beyond static responses, connecting dynamic data sources directly through .NET chat and embedding frameworks.</p>\n\n<p><strong>Agent Framework and Multi-Agent Workflows:</strong> The next segment focuses on Microsoft Agent Framework, extending Microsoft Extensions for AI <a class=\"medius-aisummary\" data-starttime=\"00:33:09\">00:33:09</a>. Bruno explains how agents are created atop chat clients by adding descriptive instructions, transforming them into autonomous entities capable of writing, editing, or generating content <a class=\"medius-aisummary\" data-starttime=\"00:35:47\">00:35:47</a>. Through examples, he shows how agents collaborate by sharing outputs sequentially in workflows, powered by built-in interfaces like IAgent and visualized using the Dev UI tool. He introduces Agent-to-Agent (A2A) protocol support for cross-agent communication, illustrating NVIDIA Nemo and Microsoft agents interacting via coordinated URLs <a class=\"medius-aisummary\" data-starttime=\"00:39:54\">00:39:54</a>. The ability to link multiple models, external APIs, and internal logic via agents and workflows demonstrates the full modular power of .NET AI integrations.</p>\n\n<p><strong>Conclusion and Resources:</strong> In closing, Bruno summarizes the AI ecosystem in .NET encompassing chat interactions, embeddings, RAG pipelines, MCP integrations, and Agent Framework orchestration <a class=\"medius-aisummary\" data-starttime=\"00:42:39\">00:42:39</a>. He revisits the showcase application where all components—data analysis, image generation, and multi-agent collaboration—work together under Aspire coordination. Developers are encouraged to explore Microsoft's official libraries and documentation at aiappsfor.net and Generative AI for Beginners in .NET for extended learning. Bruno wraps up by thanking viewers, reinforcing that these libraries empower developers to build complex AI solutions natively in C#, with direct support for both cloud and local environments <a class=\"medius-aisummary\" data-starttime=\"00:43:45\">00:43:45</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.463Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/77f11a3c-9f3d-4ac6-9ea4-fef86c97f73a","ogTitle":"AI Building Blocks for .NET: Add intelligence to your C# apps","ogDescription":"A practical, opinionated guide to building intelligent apps in .NET.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":".NET","logicalValue":".NET"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/dotnet-ai","backpackLink":null,"label":"CTA","category":"Web link"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"3c69adde-fa78-473a-9387-bb6d135c744f","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Inbal Sagiv","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777500179328001VUWe","langLocale":"en-US","sessionId":"1777500179328001VUWe","sessionInstanceId":"1777500179328001VUWe","sessionCode":"OD839","sessionCodeNormalized":"EN-US-OD839","title":"AI solutions built to power industrial innovation and sovereign control","sortTitle":"ai solutions built to power industrial innovation and sovereign control","description":"Learn how Microsoft Foundry brings AI to industrial and sovereign environments with Foundry Local on Azure Local. This session shows how organizations can build and run AI applications directly on Azure Local infrastructure with low latency, local data control, and support for connected or fully disconnected operations - while maintaining a consistent developer and governance experience through Azure Arc.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/cd7a39ba-1b09-4b0b-be50-71000e4ac33a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/cd7a39ba-1b09-4b0b-be50-71000e4ac33a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/cd7a39ba-1b09-4b0b-be50-71000e4ac33a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/cd7a39ba-1b09-4b0b-be50-71000e4ac33a","speakerIds":["1768245294609build26-1771933786485001asHR"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.39Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/cd7a39ba-1b09-4b0b-be50-71000e4ac33a","ogTitle":"AI solutions built to power industrial innovation and sovereign control","ogDescription":"Learn how Microsoft Foundry brings AI to industrial and sovereign environments with Foundry Local on Azure Local. This session shows how organizations can build and run AI applications directly on Azure Local infrastructure with low latency, local data control, and support for connected or fully disconnected operations - while maintaining a consistent developer and governance experience through Azure Arc.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1771933786485001asHR","RegistrantKey":"1768245294609build26-1771933786485001asHR"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/FoundryLocalAzure_PreviewRequest","backpackLink":null,"label":"Register to get access to Foundry Local on Azure Local preview","category":"Web link"},{"link":"https://aka.ms/build26blog","backpackLink":null,"label":"Read the blog post with related Build announcements","category":"Web link"},{"link":"https://learn.microsoft.com/en-us/azure/azure-sovereign-clouds/private/foundry-local/what-is-foundry-local-on-azure-local","backpackLink":null,"label":"Explore the product documentation:","category":"Web link"},{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aks.ms/build26/OD839","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Azure Local","logicalValue":"Azure Local"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"10b2e2ce-39d3-4c07-b155-368e3283a1f3","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Adrian Macias","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778213291088001iO8Y","langLocale":"en-US","sessionId":"1778213291088001iO8Y","sessionInstanceId":"1778213291088001iO8Y","sessionCode":"LIVESP120","sessionCodeNormalized":"EN-US-LIVESP120","title":"AMD quick take: Live from Microsoft Build","sortTitle":"amd quick take: live from microsoft build","description":"Get a sneak peek of the latest from AMD featuring Microsoft’s Karuana Gatimu and Adrian Macias","aiDescription":"<p><strong>Introduction and Event Setting:</strong> The video opens at the Microsoft Build conference in San Francisco, set against the bustling backdrop of the AMD booth on the show floor (<a class=\"medius-aisummary\" data-starttime=\"00:00:04\">00:00:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>). The host expresses enthusiasm to speak with Adrian about the innovations empowering the global developer community. Adrian joins with equal excitement, acknowledging the rapid pace of change in the technology space and the wealth of developments taking place, setting the stage for a conversation about the partnership between Microsoft and AMD (<a class=\"medius-aisummary\" data-starttime=\"00:00:21\">00:00:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>).</p>\n\n<p><strong>Microsoft and AMD Partnership:</strong> The discussion pivots to why the Microsoft-AMD collaboration matters for developers (<a class=\"medius-aisummary\" data-starttime=\"00:00:30\">00:00:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>). Adrian explains that while technical innovations are important, they are incomplete without a strong and supportive ecosystem. This partnership focuses on cultivating a thriving developer community—one that encourages creativity and supports new initiatives. The companies are collaborating to enable a shared platform where technology and community growth fuel one another, reinforcing the broader goal of empowering developers through both tools and culture (<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>).</p>\n\n<p><strong>Experimentation and Innovation:</strong> The conversation then transitions to the unprecedented pace of technological change, which Adrian describes as transformative for how professionals approach work, platforms, and experimentation (<a class=\"medius-aisummary\" data-starttime=\"00:00:48\">00:00:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:56\">00:00:56</a>). He highlights that experimentation is now more accessible than ever, empowering developers to explore choices in technologies, infrastructure, and models. This environment encourages innovation because the barriers to testing new ideas are lower and the opportunities to iterate quickly are greater. Adrian emphasizes that this experimental freedom is a defining feature of the current era of software and hardware co-innovation (<a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:09\">00:01:09</a>).</p>\n\n<p><strong>Cost Efficiency and Developer Empowerment:</strong> Continuing on the theme of experimentation, Adrian notes the significant shift in cost management for developers (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:38\">00:01:38</a>). Previously, organizations had to commit capital and resources upfront before seeing results from their technology choices. Today, with cloud platforms and scalable architectures, the cost to explore new solutions has effectively dropped to near zero. This affordability enables developers and designers to stress-test ideas and deploy solutions at lower risk. While this flexibility can be overwhelming, Adrian frames it as an empowering shift that democratizes innovation and makes advanced computing resources accessible to a wider range of creators (<a class=\"medius-aisummary\" data-starttime=\"00:01:38\">00:01:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>).</p>\n\n<p><strong>Closing Remarks and Resources:</strong> As the conversation wraps up, the host thanks Adrian and AMD for their continued partnership with Microsoft and their contributions to the developer community (<a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:57\">00:01:57</a>). To help viewers explore all the content presented at Build, the host directs them to AMD’s sponsor page on Build.Microsoft.com (<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:08\">00:02:08</a>). There, attendees can find resources, additional sessions, and materials to continue learning and experimenting. The video concludes with an encouraging sign-off inviting developers to stay engaged and discover more innovations coming from the evolving Microsoft and AMD collaboration (<a class=\"medius-aisummary\" data-starttime=\"00:02:09\">00:02:09</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:17\">00:02:17</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ff979afb-cc26-462b-8635-dbc137ad9c1a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ff979afb-cc26-462b-8635-dbc137ad9c1a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ff979afb-cc26-462b-8635-dbc137ad9c1a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ff979afb-cc26-462b-8635-dbc137ad9c1a","speakerIds":["1768245294609build26-1776355971597001QUS8"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.773Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["AMD"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ff979afb-cc26-462b-8635-dbc137ad9c1a","ogTitle":"AMD quick take: Live from Microsoft Build","ogDescription":"Get a sneak peek of the latest from AMD featuring Microsoft’s Karuana Gatimu and Adrian Macias","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"","logicalValue":""}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1776355971597001QUS8","RegistrantKey":"1768245294609build26-1776355971597001QUS8"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"cb00db36-8754-4523-92a6-a2822954c587","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Cliff Simpkins","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396132631001Fpnz","langLocale":"en-US","sessionId":"1774396132631001Fpnz","sessionInstanceId":"1774396132631001Fpnz","sessionCode":"ODSP906","sessionCodeNormalized":"EN-US-ODSP906","title":"Apply orchestration patterns for production AI agents","sortTitle":"apply orchestration patterns for production ai agents","description":"Moving your agent from demo to production requires a platform that can support and enable it. You will learn about enabling robots, humans, and agents to come together in a scalable and secure way. UiPath demos the latest innovations, including new business orchestration design canvases, native coding agent support, and expanded agentic testing capabilities. You don't have to choose between the developer experience and the power of an enterprise platform.","aiDescription":"<p><strong>Introduction to UiPath and Session Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Cliff Simpkins begins the session by welcoming viewers and introducing UiPath as Microsoft’s preferred enterprise automation platform built on Azure. He explains that the next 20 minutes will cover what UiPath does, why it matters for developers, and a highlight reel from their recent developer conference. Simpkins describes UiPath’s relevance for builders, emphasizing three developer-facing pillars: Maestro as the execution layer that makes agentic production-ready automation possible; native integration with GitHub Copilot through NPM installation for UiPath skills and CLI; and built-in Azure security and compliance inheritance for enterprise deployment. He positions UiPath as an evolution from robotic process automation (RPA) to a full orchestration runtime capable of integrating agents, robots, human reviews, and exceptions in a governed, auditable environment (<a class=\"medius-aisummary\" data-starttime=\"00:01:15\">00:01:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:04\">00:02:04</a>).</p>\n\n<p><strong>UiPath Maestro Platform and Microsoft Ecosystem Integration:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:02:06\">00:02:06</a> through <a class=\"medius-aisummary\" data-starttime=\"00:04:25\">00:04:25</a>, Simpkins describes Maestro—the orchestration engine at UiPath’s core that manages AI agents, human approvals, and system integrations across durable, observable processes. Three modeling canvases are introduced: Maestro BPMN for business workflows, Maestro Flow for developer-centric orchestration, and Maestro Case for dynamic enterprise processes. He demonstrates how UiPath fits within the Microsoft-first enterprise workflow where Teams, Outlook, SharePoint, and Copilot collaborate with UiPath’s orchestration layer to handle complex tasks like claims and approvals. He highlights two new Microsoft integrations: conversational UiPath agents that operate directly in Microsoft Teams, and GitHub Copilot gaining full UiPath context by installing its skills and CLI, enabling developers to build and deploy agentic automations seamlessly inside Visual Studio Code.</p>\n\n<p><strong>Developer Conference Highlights and Coding Agent Innovations:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:04:27\">00:04:27</a>, Simpkins transitions to UiPath’s developer conference keynote from Bengaluru. Raghu Malpani introduces how automation development has transformed—40% of code now generated by AI and 85% of developers using coding agents. The session underscores developer leverage across the software lifecycle, compressing timelines dramatically. Alexandru Roman demonstrates agent-assisted debugging using Claude Code, which autonomously investigates, hypothesizes, and repairs failed automation queues (<a class=\"medius-aisummary\" data-starttime=\"00:05:50\">00:05:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:24\">00:07:24</a>). Roman illustrates agents creating repositories, UI elements, and deploying fixes rapidly, then showcases cross-platform use by switching to Gemini CLI for queue-based automations. These demonstrations prove UiPath’s flexibility—agents can work across environments with full governance and extensibility.</p>\n\n<p><strong>Exploring Maestro Flow and Enterprise Process Modeling:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:08:12\">00:08:12</a>, Vikram Kakumani and Milo Shields introduce Maestro Flow—the new visual canvas for developers. Built atop Maestro’s enterprise-grade runtime, Flow provides a schema natively understood by coding agents for complete file manipulation. Shields demonstrates creating a billing dispute process by combining manual and Outlook triggers, classification agents, and advanced error handling, including PII and harmful content detection guardrails (<a class=\"medius-aisummary\" data-starttime=\"00:09:15\">00:09:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>). He details trace views, real-time approvals, Slack notifications, and seamless integration with Git version control. Flow’s evaluation framework enables end-to-end testing for orchestrations, blending developer enjoyment with enterprise rigor. Kakumani then moves to Maestro Case as a solution for modeling complex, non-linear business workflows like loan origination, where rule-based or agent reasoning dynamically activates various stages without predefined paths (<a class=\"medius-aisummary\" data-starttime=\"00:13:35\">00:13:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:05\">00:16:05</a>).</p>\n\n<p><strong>Agentic Testing and Autonomous QA Capabilities:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:16:12\">00:16:12</a>, Anvita Dekhane introduces UiPath’s Test Cloud—an AI-driven agentic testing suite that transforms manual QA into autonomous validation. She outlines three advancements under autonomous software testing: UiPath Autopilot for running manual tests independently; exploration mode for dynamic test coverage and case mapping; and UiPath Delegate—a digital coworker removing administrative overhead. Tester Amol Awate demonstrates Delegate executing tests at Accrual Bank, emailing managers, filing bugs in Jira, and updating test results directly in context (<a class=\"medius-aisummary\" data-starttime=\"00:17:57\">00:17:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:20\">00:19:20</a>). His workflow showcases UiPath Skills integration with applications and test tools, emphasizing a seamless, low-effort QA cycle across enterprise systems.</p>\n\n<p><strong>Closing Remarks and Developer Resources:</strong> Concluding at <a class=\"medius-aisummary\" data-starttime=\"00:19:23\">00:19:23</a>, Simpkins reviews UiPath’s expanded orchestration and coding agent ecosystem. He confirms availability of the CLI and skills with GitHub Copilot support and early access for Maestro Flow. Developers building AI agents for enterprise environments with auditability and cross-system collaboration are encouraged to begin with a simple install of the UiPath CLI and skills. The session closes with guidance to access SDKs, documentation, and community resources via uipath.com/developers, reaffirming UiPath’s mission to empower developers building secure, governed, agentic automations within familiar tools like Visual Studio Code.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/716ac10b-3253-4e07-af14-68866af36f67","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/716ac10b-3253-4e07-af14-68866af36f67","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/716ac10b-3253-4e07-af14-68866af36f67","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/716ac10b-3253-4e07-af14-68866af36f67","speakerIds":["1768245294609build26-1778081207527001zsSs"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.638Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["UiPath"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/716ac10b-3253-4e07-af14-68866af36f67","ogTitle":"Apply orchestration patterns for production AI agents","ogDescription":"Moving your agent from demo to production requires a platform that can support and enable it. You will learn about enabling robots, humans, and agents to come together in a scalable and secure way. UiPath demos the latest innovations, including new business orchestration design canvases, native coding agent support, and expanded agentic testing capabilities. You don't have to choose between the developer experience and the power of an enterprise platform.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778081207527001zsSs","RegistrantKey":"1768245294609build26-1778081207527001zsSs"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"MS Teams","logicalValue":"MS Teams"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"03020721-45bc-433d-8b93-7ca31d350c59","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Saurabh Sensharma, Saloni Sonpal, Vishnu Charan TJ","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778257839117001VWQp","langLocale":"en-US","sessionId":"1778257839117001VWQp","sessionInstanceId":"1778257839117001VWQp","sessionCode":"OD870","sessionCodeNormalized":"EN-US-OD870","title":"Azure Storage for AI workloads​","sortTitle":"azure storage for ai workloads​","description":"Learn how Azure Storage powers AI inference at scale. This session explores how to securely bring enterprise data to AI models, accelerate AI workloads with high-performance storage, and reduce GPU idle time through faster model loading and optimized data access. See how Azure Storage integrates with Microsoft and open-source AI frameworks to improve performance, lower costs, and enable scalable agent-based applications.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The presentation begins at <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a> with a welcome from Saurabh Sen Sharma, Product Manager for Azure Storage. He introduces colleagues Vishnu Charan and Saloni Sonpal and outlines the session’s focus: showcasing how Azure Storage powers AI inference at scale, ensures secure integration of enterprise data with AI models, accelerates performance through optimized storage, and reduces GPU idle time. Between <a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:09\">00:01:09</a>, the agenda emphasizes inference—the dominant AI workload today—and the capabilities Azure brings to make it faster, cheaper, and more secure. The opening frames Azure Storage both as foundational for AI workloads and as a system where AI enhances storage intelligence, before narrowing the focus to storage for AI inference.</p>\n\n<p><strong>Azure Storage Foundation for AI:</strong> Starting at <a class=\"medius-aisummary\" data-starttime=\"00:02:05\">00:02:05</a>, the session dives into how Azure Storage integrates into the AI stack across training and inference, supporting flexibility across managed platform services, Kubernetes, and infrastructure as a service. Key technologies described between <a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:17\">00:04:17</a> include Azure Blob Storage, Blobfuse, ADLFS, and Pytorch connectors delivering high throughput data access. These enable efficient workflows ranging from data preparation to model retrieval. The talk highlights orchestrations through VMSS and Azure Kubernetes Service (AKS) and mentions integration with Ray, TensorFlow, and ONNX frameworks. From <a class=\"medius-aisummary\" data-starttime=\"00:04:18\">00:04:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>, the presenters outline three main paths for AI workloads—Microsoft Foundry (managed platform), Azure Kubernetes Service, and IaaS—each leveraging storage as a critical backbone for secure and scalable data-driven AI operations.</p>\n\n<p><strong>Agentic Inference and Prompt Caching:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:05:20\">00:05:20</a>, attention shifts to agentic inference from a storage perspective, detailing two stages: data preparation and runtime management. Between <a class=\"medius-aisummary\" data-starttime=\"00:06:07\">00:06:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>, the discussion explains how Azure Blob stores embeddings and indexes, providing millisecond retrieval during inference. The session then transitions at <a class=\"medius-aisummary\" data-starttime=\"00:07:11\">00:07:11</a> to focus on optimizing inference using prompt caching—a key advancement reducing redundant GPU computations. From <a class=\"medius-aisummary\" data-starttime=\"00:08:03\">00:08:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:49\">00:09:49</a>, the speakers illustrate how caching reused attention states dramatically cuts GPU cycles, lowers latency, and slashes inference costs up to 90%. They then introduce Microsoft’s three tiers of caching—implicit, extended, and explicit—between <a class=\"medius-aisummary\" data-starttime=\"00:10:10\">00:10:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:03\">00:13:03</a>, culminating in Azure Context Cache. Built on Azure Blob Storage, this tier persists context across sessions, nodes, and regions with TTLs extending weeks, adding enterprise-grade durability and portability.</p>\n\n<p><strong>Azure Context Cache Design and Developer Benefits:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:13:05\">00:13:05</a>, the architecture of Azure Context Cache is explained—offering encrypted provider-only endpoints, enterprise-controlled retention settings, and performance achieving millions of tokens per minute. From <a class=\"medius-aisummary\" data-starttime=\"00:14:19\">00:14:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:12\">00:16:12</a>, the conversation highlights the system’s impact: improving cost efficiency, throughput, and latency while preserving model intellectual property and observability. The explicit caching feature enters gated preview at Microsoft Build 2026, inviting developers to test long-context agent scenarios. Vishnu Charan then takes over at <a class=\"medius-aisummary\" data-starttime=\"00:16:14\">00:16:14</a> to demonstrate how open-source frameworks leverage the same caching through integrations such as Nvidia’s NVCEL library and LM Cache, which offload KV states directly to Azure Blob. A live demo between <a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:41\">00:19:41</a> shows model load times cut from 52 seconds to 8 seconds, proving durable caching accelerates inference dramatically.</p>\n\n<p><strong>Model Loading and Distributed Caching:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:20:00\">00:20:00</a> onward, Vishnu showcases the Azure Blob plug-in for “RunEI Model Streamer,” allowing parallel chunk loading directly from object storage. Demonstrations at <a class=\"medius-aisummary\" data-starttime=\"00:21:45\">00:21:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:08\">00:24:08</a> reveal GPUs reaching full network saturation, achieving 3–6 times faster cold starts than default loaders. He further extends this to distributed AKS clusters at <a class=\"medius-aisummary\" data-starttime=\"00:24:45\">00:24:45</a>, introducing Azure Container Storage distributed cache—a feature enabling multi-node parallel reads with NVMe-class latency, eliminating redundant downloads, and reducing multi-node startup times by 2.5x. Early access signups are invited for teams running AI on Kubernetes. The section confirms Azure’s ability to maintain high utilization across inference compute layers through optimized storage architecture.</p>\n\n<p><strong>Bringing Enterprise Data into AI and Conclusion:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:25:05\">00:25:05</a>, focus shifts to connecting enterprise data with AI models. The speakers explain integrations like Blobfuse, ADLFS, PyTorch connectors, and LangChain Azure Blob loaders, collectively enabling frictionless, high-speed access to data for inference and retrieval pipelines. At <a class=\"medius-aisummary\" data-starttime=\"00:27:17\">00:27:17</a>, Saurabh introduces Azure Files as a source for Azure AI Search and details new connectors that streamline transforming file data into RAG-ready indexes. Foundry IQ is unveiled at <a class=\"medius-aisummary\" data-starttime=\"00:29:07\">00:29:07</a> as Microsoft’s unified knowledge layer, simplifying enterprise data grounding for agents with built-in security and governance via Azure Blob. Finally, Azure Storage Center, introduced around <a class=\"medius-aisummary\" data-starttime=\"00:30:23\">00:30:23</a>, provides a centralized hub uniting object, file, and block storage—with preview signups open. The session concludes at <a class=\"medius-aisummary\" data-starttime=\"00:32:31\">00:32:31</a>, reaffirming Azure Storage’s delivery against three key requirements: enterprise data access, fast model loading, and cost-efficient infrastructure utilization. Together, these advancements empower developers to build scalable, secure, and high-performing AI applications on Azure.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e68dfeb1-b10a-463f-b2d8-53707101babe","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e68dfeb1-b10a-463f-b2d8-53707101babe","captionFileLink":"https://medius.microsoft.com/video/asset/Caption/e68dfeb1-b10a-463f-b2d8-53707101babe","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e68dfeb1-b10a-463f-b2d8-53707101babe","speakerIds":["1768245294609build26-1778257946075001QaMe","1768245294609build26-1778257989205001rfcu","1768245294609build26-1778258032365001A3Wu"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.774Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e68dfeb1-b10a-463f-b2d8-53707101babe","ogTitle":"Azure Storage for AI workloads​","ogDescription":"Learn how Azure Storage powers AI inference at scale. This session explores how to securely bring enterprise data to AI models, accelerate AI workloads with high-performance storage, and reduce GPU idle time through faster model loading and optimized data access. See how Azure Storage integrates with Microsoft and open-source AI frameworks to improve performance, lower costs, and enable scalable agent-based applications.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778257946075001QaMe","RegistrantKey":"1768245294609build26-1778257946075001QaMe"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778258032365001A3Wu","RegistrantKey":"1768245294609build26-1778258032365001A3Wu"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778257989205001rfcu","RegistrantKey":"1768245294609build26-1778257989205001rfcu"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"bd8a275d-3da2-442c-b0dd-68c62c98b945","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396534856001z2g8","langLocale":"en-US","sessionId":"1774396534856001z2g8","sessionInstanceId":"1774396534856001z2g8","sessionCode":"ODSP912","sessionCodeNormalized":"EN-US-ODSP912","title":"Build agentic testing systems to validate AI generated code","sortTitle":"build agentic testing systems to validate ai generated code","description":"AI is driving rapid expansion of software, but vibe coding lacks the rigor needed for reliability. Traditional tests can’t keep pace with agent-generated logic. To scale, we must move beyond manual checks to Agentic Testing: using autonomous agents to validate autonomous systems. Explore patterns for creating autonomous test agents, detecting failures across workflows, and continuously verifying behavior. Take away practical approaches to make AI-driven software reliable and production ready.","aiDescription":"<p><strong>Introduction and Context:</strong> The session begins with a greeting at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a> as the speaker, a Developer Relations Manager at Tesma AI, welcomes everyone to Microsoft Build 2D6 and introduces the focus of the talk—agent testing. The discussion frames how modern AI agents can now write, run, break, and patch code, but emphasizes that their ability to test code reliably before production remains limited. The real challenge, as stated at <a class=\"medius-aisummary\" data-starttime=\"00:00:30\">00:00:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:40\">00:00:40</a>, is not speed but trust—ensuring that testing performed by autonomous agents can be verified and relied upon without human intuition.</p>\n\n<p><strong>The Testing Gap and Introduction of Keynes CLI:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:11\">00:01:11</a>, the speaker introduces the solution: Keynes CLI, Tesma AI’s new command-line tool designed to close the validation gap in the agent testing process. This product allows agents to perform testing like a careful engineer, ensuring reliable outputs. The talk underscores that the focus should shift from how fast we build code to how effectively we validate it. Keynes CLI acts as a deterministic validation layer between development and shipping, giving developers and AI systems a way to verify applications originally built by autonomous agents.</p>\n\n<p><strong>Capabilities and Core Features of Keynes CLI:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>, the presentation dives into the capabilities of Keynes CLI. It integrates natively with various agent systems such as Cloud Codex, Gemini, and Copilot, allowing tests to be described using natural language prompts. The tool automatically runs validation across browsers, devices, and environments, enabling developers to release features confidently. Out-of-the-box functionality includes intent-based control without the need for manual test scripts, XPaths, or selectors; resilient runs that persist through complex journeys; and built-in bug discovery using vision-based waits. It automatically generates Playwright test scripts from natural language inputs and stores test cases in markdown format that can be replayed or integrated into CI/CD pipelines. Keynes CLI also provides auto-heal functionality to update test cases after UI changes and outputs structured ND JSON artifacts readable by agents.</p>\n\n<p><strong>Usage Options and Integration Examples:</strong> By <a class=\"medius-aisummary\" data-starttime=\"00:04:28\">00:04:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>, the speaker explains three key methods to use Keynes CLI: directly from the command-line interface, through an SDK for CI/CD pipelines or custom agents, and by enabling agent integration using the provided agent MD file. A demonstration follows showing how the tool can be installed from NPM and used in interactive mode to execute objectives. The live demo showcases automating tasks such as completing a checkout flow and pasting an order ID into Google, with Keynes CLI autonomously dividing the tasks into workflows and performing them step-by-step using a local browser interface.</p>\n\n<p><strong>Demonstration and Live Workflow Execution:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:10\">00:11:10</a>, the session illustrates Keynes CLI executing an automated workflow. It completes the order placement, extracts the order ID, and successfully performs subsequent tasks like inserting that ID into a Google search. Each action is visible in real-time terminal logs and browser interactions. After test completion, users can exit and save sessions, review generated Playwright scripts, and access shareable log links that contain the full trace and visual evidence of the test’s steps. These outputs allow validation before merging features, and any detected bugs are automatically flagged for review.</p>\n\n<p><strong>Agent Collaboration and Closing Insights:</strong> In the final portion beginning around <a class=\"medius-aisummary\" data-starttime=\"00:12:36\">00:12:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:03\">00:13:03</a>, the speaker demonstrates Keynes CLI’s ability to operate collaboratively with AI agents in headless mode, producing results, completion statuses, and shareable proof logs. The demo underscores how agents equipped with Keynes CLI can autonomously generate, execute, and verify end-to-end test cases during development. The tool’s structured outputs and validation proofs help developers build trusted applications in the agent-driven era, ultimately bridging the trust gap between automation and confident product release.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d4a99675-b19b-47d1-81df-c85d93167c06","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d4a99675-b19b-47d1-81df-c85d93167c06","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d4a99675-b19b-47d1-81df-c85d93167c06","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d4a99675-b19b-47d1-81df-c85d93167c06","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.64Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["TestMu AI (formerly LambdaTest)"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d4a99675-b19b-47d1-81df-c85d93167c06","ogTitle":"Build agentic testing systems to validate AI generated code","ogDescription":"AI is driving rapid expansion of software, but vibe coding lacks the rigor needed for reliability. Traditional tests can’t keep pace with agent-generated logic. To scale, we must move beyond manual checks to Agentic Testing: using autonomous agents to validate autonomous systems. Explore patterns for creating autonomous test agents, detecting failures across workflows, and continuously verifying behavior. Take away practical approaches to make AI-driven software reliable and production ready.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/sess/1774396534856001z2g8/supmat/Agentic%20Testing%20%26%20The%20Future%20of%20Trust%20by%20Mohit%20Juneja_1777941467542001Zcn5.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Monitor","logicalValue":"Monitor"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"3aa3861e-7e5a-4cf6-bbc1-29c8d2f89cfa","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Paul Usher","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396471341001jVyX","langLocale":"en-US","sessionId":"1774396471341001jVyX","sessionInstanceId":"1774396471341001jVyX","sessionCode":"ODSP911","sessionCodeNormalized":"EN-US-ODSP911","title":"Build AI-first business apps that turn dashboards into actions","sortTitle":"build ai-first business apps that turn dashboards into actions","description":"Business apps are shifting from dashboards and filters to AI-driven experiences. In this session, we show how users can ask questions, get insights, and take action instantly. Using DevExpress components, we demonstrate how structured UI brings AI responses to life, turning data into clear, actionable outcomes.","aiDescription":"<p><strong>Introduction and Overview:</strong> At the beginning of the session (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>), Paul Usher from DevExpress introduces the Microsoft Build 2026 presentation, emphasizing that the focus is not on generating code or simply integrating a chatbot, but on embedding AI deeply into Blazor applications to enhance natural user experiences. He outlines the demo environment—built with Visual Studio, .NET, DevExpress Blazor controls, DevExpress Reporting, and Azure OpenAI connected through an iChat client. The session explores three practical use cases: AI-powered DevExpress grid interactions via tool calling, AI-integrated translations within a report viewer, and Azure OpenAI-assisted contract analysis visualized through DevExpress reports.</p>\n\n<p><strong>AI-Driven Data Interaction in Blazor Grids:</strong> The demonstration begins with the main dashboard (<a class=\"medius-aisummary\" data-starttime=\"00:01:03\">00:01:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>) featuring KPI cards, charts, and a data grid tied to 10,000 rows of sales information. Initially, Paul manually filters and sorts the grid to show conventional interactions. Then he introduces the DX AI Chat control, which allows users to use natural language for data operations—such as grouping orders by customer, filtering to Texas, or exporting results to Excel. Rather than simply responding with text, the AI triggers DevExpress grid API calls like “filterRiskAccounts,” dynamically changing the grid. This collaboration demonstrates how AI processes intent while DevExpress handles structured interaction, transforming chat responses into direct application actions.</p>\n\n<p><strong>Technical Breakdown and Tool Calling Integration:</strong> Transitioning to Visual Studio (<a class=\"medius-aisummary\" data-starttime=\"00:04:05\">00:04:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:33\">00:07:33</a>), Paul dissects how the integration is structured. The DX AI chat control within Home.razor is linked to a “DX tools” AI client, defined in Program.cs and configured with Azure OpenAI details sourced from .NET user secrets. The AI Tools Context Builder is used to register allowed capabilities—methods like “filterByRegion,” “export,” or “summarize.” These methods form a toolbox that defines what the AI model can invoke. The “filterByRegion” example shows metadata annotations that help the model understand when and how to use each method. Importantly, live DevExpress component instances are injected at runtime, allowing AI to execute real API functions while maintaining strict control over permitted actions. This creates a secure bridge where user intent is translated into controlled UI behavior.</p>\n\n<p><strong>AI Integration in Reporting and Translation:</strong> The next section (<a class=\"medius-aisummary\" data-starttime=\"00:07:39\">00:07:39</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:03\">00:09:03</a>) shifts to the DevExpress report viewer, highlighting built-in AI translation features. A quarterly memo, initially in French, is automatically translated into English using AI-driven reporting extensions. This translation workflow requires no separate chat UI—the report viewer itself manages the AI process. Within Visual Studio, Paul reveals how this functionality is enabled via “AddBlazorReportingAIIntegration” calls in Program.cs. The translation pipeline supports multiple languages and integrates inline translation rendering so users can interact directly with translated documents. Here, the AI is fully embedded within the control, reinforcing the model of self-contained, domain-specific intelligence where the DevExpress component dictates how AI enhances user tasks.</p>\n\n<p><strong>Contract Review Workflow and Custom AI Application:</strong> In the final major demo (<a class=\"medius-aisummary\" data-starttime=\"00:09:12\">00:09:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:43\">00:11:43</a>), the focus moves to contract analysis. Within the DevExpress report viewer, Azure OpenAI reviews a master service agreement, identifying risky or one-sided clauses that are then visually marked inside the document with warning tags and highlight borders. The “ReviewAsync” method injects the iChat client directly, sending the full contract text and receiving a structured response outlining clause risks. The AI results feed into the report, modifying its data source so each clause is tagged as risky or safe. The key insight is that the report itself remains AI-agnostic—it simply reacts to a data field state. Thus, AI produces analytical context while DevExpress visualizes results through its reporting engine.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> Wrapping up (<a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:31\">00:13:31</a>), Paul emphasizes that these three pages exemplify complementary AI integration models: tool calling within grids, embedded AI processing within reports, and full workflow composition for document review. Each approach illustrates how AI and DevExpress components cooperate effectively—the AI interprets user intent, and the controls deliver structured outcomes. The architectural pattern deliberately keeps AI as a supporting layer rather than an overarching system, preserving clear separation between UI, data services, and intelligent function calls. The session concludes noting that Azure OpenAI provides intelligence, DevExpress delivers interaction frameworks, and Visual Studio unifies the ecosystem for intelligent, user-centered applications.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8041cdaf-f3df-44ea-aa33-0dd7e1dd2745","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8041cdaf-f3df-44ea-aa33-0dd7e1dd2745","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8041cdaf-f3df-44ea-aa33-0dd7e1dd2745","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8041cdaf-f3df-44ea-aa33-0dd7e1dd2745","speakerIds":["1768245294609build26-1776365404890001SeT6"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.64Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["DevExpress"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8041cdaf-f3df-44ea-aa33-0dd7e1dd2745","ogTitle":"Build AI-first business apps that turn dashboards into actions","ogDescription":"Business apps are shifting from dashboards and filters to AI-driven experiences. In this session, we show how users can ask questions, get insights, and take action instantly. Using DevExpress components, we demonstrate how structured UI brings AI responses to life, turning data into clear, actionable outcomes.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776365404890001SeT6","RegistrantKey":"1768245294609build26-1776365404890001SeT6"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Software Development Company","logicalValue":"Software Development Company"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"47ad3060-77a9-497c-a407-6c9901adbef8","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396027120001rQUA","langLocale":"en-US","sessionId":"1774396027120001rQUA","sessionInstanceId":"1774396027120001rQUA","sessionCode":"ODSP904","sessionCodeNormalized":"EN-US-ODSP904","title":"Build AI-powered collaboration on a visual canvas","sortTitle":"build ai-powered collaboration on a visual canvas","description":"Visual collaboration is evolving as AI becomes part of the workflow—not just the output.\nSee how to go from prompt to diagram in seconds, generate and organize boards with AI, and connect AI agents to the Lucid MCP server to create and update documents programmatically.\nLearn how to build AI-driven collaboration workflows on a shared canvas, making work visible, structured, and automatable across teams and tools.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/161cb593-dbf4-471c-bd53-aeba3bdb58ee","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/161cb593-dbf4-471c-bd53-aeba3bdb58ee","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/161cb593-dbf4-471c-bd53-aeba3bdb58ee","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/161cb593-dbf4-471c-bd53-aeba3bdb58ee","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.638Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Lucid Software"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/161cb593-dbf4-471c-bd53-aeba3bdb58ee","ogTitle":"Build AI-powered collaboration on a visual canvas","ogDescription":"Visual collaboration is evolving as AI becomes part of the workflow—not just the output.\nSee how to go from prompt to diagram in seconds, generate and organize boards with AI, and connect AI agents to the Lucid MCP server to create and update documents programmatically.\nLearn how to build AI-driven collaboration workflows on a shared canvas, making work visible, structured, and automatable across teams and tools.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"}]},{"@search.score":1.0,"documentId":"d25d9352-e98c-4141-a271-417ea5a59262","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Shriram Sankaran, Vishnu Menon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774395897516001yIE4","langLocale":"en-US","sessionId":"1774395897516001yIE4","sessionInstanceId":"1774395897516001yIE4","sessionCode":"ODSP902","sessionCodeNormalized":"EN-US-ODSP902","title":"Build AI‑driven UIs in .NET MAUI with design systems","sortTitle":"build ai‑driven uis in .net maui with design systems","description":"This session shows how to build a real‑world .NET MAUI application using structured design systems and AI‑assisted UI composition. Start by defining a design system and translating it into reusable MAUI components and tokens. Learn how AI‑driven UI workflows help compose screens, apply theming, and enforce consistency across the app—accelerating UI development without losing control, performance, or maintainability.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Vishnu Menon, Senior Product Manager at Syncfusion, greets the audience and introduces the session topic—building AI-driven user interfaces in .NET MAUI applications. He explains that the goal is to make UIs faster, more consistent, and intelligent by leveraging design systems. He outlines the agenda at <a class=\"medius-aisummary\" data-starttime=\"00:00:20\">00:00:20</a>, which covers challenges in UI building, why AI struggles in this area, an exploration of design system standards, AI skills, and an AI-driven UI workflow, followed by a demo showcasing these ideas in action.</p>\n\n<p><strong>Demonstrating Basic AI-Generated UI and Its Limitations:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>, Vishnu prompts an AI tool to create a simple employee dashboard mobile application using .NET MAUI. The AI generates a functional and colorful interface, but upon closer inspection at <a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>, he notices misaligned icons and inconsistent colors. He explains that although AI can produce layouts rapidly, these UIs often lack polish and are not production-ready because AI lacks design context. At <a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>, Vishnu stresses that without understanding a project’s design system—spacing rules, patterns, and component standards—AI typically outputs guess-based implementations instead of intentional designs.</p>\n\n<p><strong>Understanding the Design System:</strong> Moving into the solution space at <a class=\"medius-aisummary\" data-starttime=\"00:02:46\">00:02:46</a>, Vishnu identifies two crucial concepts for improving AI-generated UI: the design system and skills. He elaborates at <a class=\"medius-aisummary\" data-starttime=\"00:02:58\">00:02:58</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:11\">00:03:11</a> that a design system ensures consistency across colors, typography, spacing, and reusable components. He showcases a .NET MAUI example using the \"styles.xaml\" file at <a class=\"medius-aisummary\" data-starttime=\"00:03:18\">00:03:18</a>, which defines styles for controls. By referencing shared elements like fonts or primary colors, developers avoid redundant hardcoding and repeated UI decisions. Vishnu emphasizes that adhering to a unified design system accelerates development and maintains visual harmony across enterprise applications.</p>\n\n<p><strong>Explaining Skills and Their Role in AI Design Guidance:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:04:56\">00:04:56</a>, the presentation transitions to AI “skills.” Vishnu defines a skill as a structure that guides AI through preferred instructions and rules. Demonstrating with a file at <a class=\"medius-aisummary\" data-starttime=\"00:05:17\">00:05:17</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:44\">00:05:44</a>, he explains that a skills.md file for a Syncfusion MAUI Button helps an agent understand how and when to use a control, while troubleshooting files provide recovery guidance. He clarifies at <a class=\"medius-aisummary\" data-starttime=\"00:06:03\">00:06:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a> that skills do not render the UI—they constrain decision-making—teaching AI how good design should look. The difference between outputs “with skills” and “without skills,” highlighted at <a class=\"medius-aisummary\" data-starttime=\"00:06:23\">00:06:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:48\">00:06:48</a>, reveals that an AI guided by skills creates more consistent, design-aware interfaces.</p>\n\n<p><strong>Demo: Applying Skills and Design System in Practice:</strong> The hands-on demo begins around <a class=\"medius-aisummary\" data-starttime=\"00:06:52\">00:06:52</a>. Vishnu demonstrates setting up Syncfusion’s MAUI controls and inserting skills files via the Terminal. At <a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:48\">00:07:48</a>, he shows that multiple skills are added for various controls and introduces an additional “design-system” skill for this demo. After prompting AI to use these skills at <a class=\"medius-aisummary\" data-starttime=\"00:08:08\">00:08:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:10\">00:09:10</a>, the generated dashboard includes defined models, design tokens, and a properly structured page. When he runs the application at <a class=\"medius-aisummary\" data-starttime=\"00:09:14\">00:09:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:29\">00:09:29</a>, the resulting UI is well-aligned, visually consistent, and reflective of the design guidelines.</p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> Wrapping up at <a class=\"medius-aisummary\" data-starttime=\"00:09:34\">00:09:34</a>, Vishnu summarizes the core message: AI should not be told to design UIs independently but instead taught how humans design them through design systems and structured skills. This guidance ensures that AI creates interfaces with the same principles and standards as expert developers. He concludes with a QR link to access the demo resources at <a class=\"medius-aisummary\" data-starttime=\"00:09:54\">00:09:54</a> and thanks the audience for attending, encouraging them to embrace “happy coding with AI.”</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d2ae05f2-1269-4f1c-aaa0-888f9b8d4e7a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d2ae05f2-1269-4f1c-aaa0-888f9b8d4e7a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d2ae05f2-1269-4f1c-aaa0-888f9b8d4e7a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d2ae05f2-1269-4f1c-aaa0-888f9b8d4e7a","speakerIds":["1768245294609build26-1774530974341001La1w","1768245294609build26-1777480438617001m2Kd"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.639Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Syncfusion"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d2ae05f2-1269-4f1c-aaa0-888f9b8d4e7a","ogTitle":"Build AI‑driven UIs in .NET MAUI with design systems","ogDescription":"This session shows how to build a real‑world .NET MAUI application using structured design systems and AI‑assisted UI composition. Start by defining a design system and translating it into reusable MAUI components and tokens. Learn how AI‑driven UI workflows help compose screens, apply theming, and enforce consistency across the app—accelerating UI development without losing control, performance, or maintainability.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(400) Expert","logicalValue":"(400) Expert"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker In-person","SpeakerId":"1768245294609build26-1774530974341001La1w","RegistrantKey":"1768245294609build26-1774530974341001La1w"},{"SpeakerType":"Owner, Speaker Digital","SpeakerId":"1768245294609build26-1777480438617001m2Kd","RegistrantKey":"1768245294609build26-1777480438617001m2Kd"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"}]},{"@search.score":1.0,"documentId":"06a10d2e-e884-47dc-82ca-627d611e6587","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Roycey Cheeran, Cosmos Darwin","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777499054358001sLll","langLocale":"en-US","sessionId":"1777499054358001sLll","sessionInstanceId":"1777499054358001sLll","sessionCode":"OD837","sessionCodeNormalized":"EN-US-OD837","title":"Build and deploy AI at the edge for real-world impact","sortTitle":"build and deploy ai at the edge for real-world impact","description":"What happens when AI leaves the datacenter and enters the physical world? Experience what's next as we unveil the public preview of Azure Local on small form factor devices with Foundry Local built in. Learn how Azure Local brings cloud-consistent infrastructure to compact edge hardware, and how Foundry Local delivers AI models where latency and connectivity demand it — on the factory floor, in the field, and wherever humans and robots work side by side — no round trip to the cloud required. Walk away with a clear blueprint for shipping smarter AI systems to the edge using Azure Local and Foundry Local.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ac47412c-dbdf-499a-be26-bdf0f518bf15","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ac47412c-dbdf-499a-be26-bdf0f518bf15","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ac47412c-dbdf-499a-be26-bdf0f518bf15","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ac47412c-dbdf-499a-be26-bdf0f518bf15","speakerIds":["1768245294609build26-1776969075880001RNkj","1768245294609build26-1778596266423001NzzS"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.389Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ac47412c-dbdf-499a-be26-bdf0f518bf15","ogTitle":"Build and deploy AI at the edge for real-world impact","ogDescription":"What happens when AI leaves the datacenter and enters the physical world? Experience what's next as we unveil the public preview of Azure Local on small form factor devices with Foundry Local built in. Learn how Azure Local brings cloud-consistent infrastructure to compact edge hardware, and how Foundry Local delivers AI models where latency and connectivity demand it — on the factory floor, in the field, and wherever humans and robots work side by side — no round trip to the cloud required. Walk away with a clear blueprint for shipping smarter AI systems to the edge using Azure Local and Foundry Local.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776969075880001RNkj","RegistrantKey":"1768245294609build26-1776969075880001RNkj"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778596266423001NzzS","RegistrantKey":"1768245294609build26-1778596266423001NzzS"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"Azure Local","logicalValue":"Azure Local"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"f0fd9f5f-d27e-40eb-ac78-76b7db1cf4c4","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ziv Navoth, Edo Segal, Gillian Sheldon","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775841416367001DUGD","langLocale":"en-US","sessionId":"1775841416367001DUGD","sessionInstanceId":"1775841416367001DUGD","sessionCode":"ODSP926","sessionCodeNormalized":"EN-US-ODSP926","title":"Build collaborative agents into apps with APIs","sortTitle":"build collaborative agents into apps with apis","description":"PwC research found the top barrier to AI agent impact isn't the technology; it's connecting agents across applications and workflows to truly operate like coworkers. Most organizations deploy agents in isolation, then wonder why value stalls. This session explores how APIs serve as the relationship layer between AI models and encourage meaningful adoption, using Napster's Omniagent API as a working example of how to move agents from standalone tools to embedded, cross-functional collaborators.","aiDescription":"<p><strong>Introduction and Context:</strong> At the beginning of the session (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>), Edo Segal welcomes attendees to the Napster Companion API presentation, highlighting its purpose for developers interested in integrating video multimodal agents into their applications. He explains that previous challenges around deploying such intelligent, interactive systems stemmed largely from cost and complexity. After three years of optimization, the team has engineered a solution that drastically lowers operational costs while making deployment simple — so developers can add a video-based, human-like AI layer to their apps in just a day. Segal emphasizes that this innovation enhances user experiences and human connections at scale and notes that leading companies are already using these agents.</p>\n\n<p><strong>Technical Architecture and Cost Efficiency:</strong> Ziv Navoth takes over (<a class=\"medius-aisummary\" data-starttime=\"00:00:51\">00:00:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>) and first addresses the critical issue of affordability. He reveals that the Omniagent API runs at merely one cent per render minute when developers use their own large language models — about twenty times cheaper than competing solutions. This enables real-time multimodal video agents to become viable production tools instead of costly prototypes. Navoth then outlines the Azure-native infrastructure supporting the API: an SDK using HTTPS and WebRTC via Azure’s Front Door load balancer, AKS-managed core pods, private endpoints, and scalable orchestration layers that connect to custom or provided Azure OpenAI deployments. Key benefits include single-bill Azure integration, retention of standard governance and security policies, and flexible support for multiple LLM providers. He highlights Siemens as a major adopter, with field technicians using voice and video interactions for equipment servicing and similar setups empowering training and customer support applications.</p>\n\n<p><strong>Agent Creation and Configuration:</strong> The next section (<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>) dives into the build process. Navoth explains that provisioning an Omniagent is straightforward — done directly from the Azure portal with no separate billing. Using the dashboard or REST API, developers can configure agents by assembling four modular elements: the Omniagent persona (face, voice, personality), knowledge (data and documents), FAQs (consistent responses), and tools (actions triggered on behalf of users). Together, these form the agent’s complete intelligence. Two creation options are shown: “New Omniagent” for a functional role-based digital persona or “New Digital Twin” for recreating a specific person’s likeness, voice, and style. Users can start from Napster’s library of stock personas or build entirely new characters. In this session, they demonstrate building “Vera,” a field service specialist, adding her photo and initializing her digital avatar before linking her datasets and operational assets.</p>\n\n<p><strong>Live Example and Demonstration:</strong> After defining Vera, the walkthrough continues (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:11\">00:07:11</a>) with live configuration steps. Knowledge bases containing manuals and diagnostic data are attached, FAQs for safety and service routines are extended, and new tools like customer lookup are added. The agent is then tested in the playground environment, where Vera interacts naturally with Marcus, a field technician. When Marcus reports compressor issues, Vera recalls prior service history, identifies a recurring high-pressure switch problem, and autonomously orders replacement parts and schedules follow-up inspection. This realistic demo illustrates Omniagent’s multimodal capacity — voice, video, context-aware reasoning, and integration with backend tools — functioning seamlessly within a dynamic workflow.</p>\n\n<p><strong>Deployment Across Channels:</strong> Once testing completes (<a class=\"medius-aisummary\" data-starttime=\"00:07:11\">00:07:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:49\">00:08:49</a>), Navoth details the deployment capabilities. Each agent operates across multiple user channels such as web, mobile, phone, and text. On the web, agents render through WebRTC to appear as embedded video concierges or sales assistants; in native apps, communication occurs via low-latency WebSocket voice streaming; and for telephony, the same logic handles SIP and VoIP calls, allowing continuous coverage — even after hours. Text-based exchanges integrate through messaging APIs with shared memory, meaning user sessions persist across mediums. Conversations begun on a website can continue seamlessly via phone or chat without users restating context. This unified continuity represents the Omniagent’s “Omni” characteristic — consistent behavior and memory synchronization wherever customer engagement occurs.</p>\n\n<p><strong>Monitoring, Insights, and Closing Notes:</strong> The final segment (<a class=\"medius-aisummary\" data-starttime=\"00:08:49\">00:08:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:59\">00:09:59</a>) introduces monitoring and auditing procedures. Every interaction across all channels is logged for inspection, showing transcripts, tool activity, memory events, and user identity — enabling full traceability and analytics. Teams can measure outcomes like resolved requests or executed actions, refine prompts, and redeploy improvements. Navoth concludes by summarizing three key takeaways: first, Omniagent is available now through Azure and can be spun up from the portal; second, the quickstart enables a running agent within 15 minutes using existing Azure OpenAI resources; and third, developers attending Microsoft Build can visit the team’s booth to co-create a “digital co-worker” in person. The video closes emphasizing the simplicity, scalability, and production-grade readiness of multimodal AI agents within modern enterprise environments.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/fea669e6-5140-4ea8-b097-a765d61311ee","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/fea669e6-5140-4ea8-b097-a765d61311ee","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/fea669e6-5140-4ea8-b097-a765d61311ee","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/fea669e6-5140-4ea8-b097-a765d61311ee","speakerIds":["1768245294609build26-1777401140686001x1Tn","1768245294609build26-1776448795539001JeI6","1768245294609build26-1776448778712001lItD"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.901Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Napster"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/fea669e6-5140-4ea8-b097-a765d61311ee","ogTitle":"Build collaborative agents into apps with APIs","ogDescription":"PwC research found the top barrier to AI agent impact isn't the technology; it's connecting agents across applications and workflows to truly operate like coworkers. Most organizations deploy agents in isolation, then wonder why value stalls. This session explores how APIs serve as the relationship layer between AI models and encourage meaningful adoption, using Napster's Omniagent API as a working example of how to move agents from standalone tools to embedded, cross-functional collaborators.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Owner","SpeakerId":"1768245294609build26-1776448778712001lItD","RegistrantKey":"1768245294609build26-1776448778712001lItD"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776448795539001JeI6","RegistrantKey":"1768245294609build26-1776448795539001JeI6"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777401140686001x1Tn","RegistrantKey":"1768245294609build26-1777401140686001x1Tn"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"50ea22de-131b-4612-a7df-69ed66baacbd","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yohan Lasorsa","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774274078845001rWQf","langLocale":"en-US","sessionId":"1774274078845001rWQf","sessionInstanceId":"1774274078845001rWQf","sessionCode":"OD810","sessionCodeNormalized":"EN-US-OD810","title":"Build fast, not fragile with Rayfin and Microsoft Fabric","sortTitle":"build fast, not fragile with rayfin and microsoft fabric","description":"Ship new apps in minutes without hand-rolling auth, data, and hosting. In this session, see how RayFin on Microsoft Fabric gives developers and coding agents a code-first backend with typesafe schemas, APIs, functions, storage, and app hosting. We'll walk through going from idea to working app, with Fabric-native data ready for governance, analytics, and AI from day one.","aiDescription":"<p><strong>Introduction and Developer Challenge:</strong> The session opens with Yohan Lasorsa welcoming viewers and introducing the main theme — how to safely push AI-assisted, or \"vibe-coded,\" applications to production <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>. Lasorsa frames the discussion by noting how coding agents have dramatically improved in the past year, accelerating app development to unprecedented speeds. However, he stresses that while prototyping is now faster than ever, treating the first working demo as a production-ready foundation is a critical mistake <a class=\"medius-aisummary\" data-starttime=\"00:00:41\">00:00:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:47\">00:00:47</a>. Many organizations fall into the trap of prematurely deploying these prototypes, which often leads to outages and data breaches. Lasorsa cites surveys showing that up to 90% of AI-generated prototypes never reach production, emphasizing that while speed has evolved, production standards such as security, governance, compliance, and scalability remain constant <a class=\"medius-aisummary\" data-starttime=\"00:01:22\">00:01:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a>. </p>\n\n<p><strong>Introducing Rayfin and Session Roadmap:</strong> The talk transitions to the introduction of Rayfin — a managed backend-as-a-service built on Microsoft Fabric, designed to bridge the gap between prototype and production <a class=\"medius-aisummary\" data-starttime=\"00:02:48\">00:02:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:03\">00:03:03</a>. Unlike similar platforms, Rayfin is Fabric-native and enterprise-ready, offering managed databases, authentication, and real-time hosting with compliance and integration baked in. Lasorsa highlights that Rayfin was created with coding agents in mind to ensure AI-generated applications can instantly take advantage of enterprise-grade security and data governance <a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>. He outlines the session agenda: a series of live demos showing how to create an app with Rayfin, manage authentication and data integrity, deploy it to production, and build an intelligent analytics dashboard <a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:34\">00:04:34</a>. </p>\n\n<p><strong>Building the “Contoso Chef” App:</strong> Using VS Code, Lasorsa demonstrates creating a new project with Rayfin CLI and setting up a recipe-sharing app called “Contoso Chef” <a class=\"medius-aisummary\" data-starttime=\"00:04:45\">00:04:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:02\">00:06:02</a>. The app idea stems from his family’s Italian cooking traditions — preserving private and public recipes digitally, with strong access control. The demo shows how Copilot CLI uses Rayfin skills to scaffold the project, generate a working schema, and connect to Microsoft Fabric as its backend <a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>. When running the app locally, Rayfin automatically generates secured APIs, authentication via Fabric OAuth, and type-safe data models for entities like Recipe and Like. This section illustrates how coding agents can focus on implementing features rather than reinventing architecture or database logic <a class=\"medius-aisummary\" data-starttime=\"00:12:03\">00:12:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:15\">00:14:15</a>. </p>\n\n<p><strong>Testing Access Control and Data Governance:</strong> In the next demonstration, Lasorsa verifies that Rayfin enforces privacy and access rules correctly by checking whether private recipes remain inaccessible to unauthenticated users <a class=\"medius-aisummary\" data-starttime=\"00:16:33\">00:16:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:30\">00:17:30</a>. He tests multiple visibility settings — private, unlisted, and public — and confirms that Rayfin’s policy-driven backend enforces these distinctions automatically. Inspecting the entity’s read policies demonstrates that only authenticated users who own a recipe can modify or view private content <a class=\"medius-aisummary\" data-starttime=\"00:20:01\">00:20:01</a>. To illustrate flexible extensibility, Lasorsa adds a new “Comment” entity using Copilot, then applies Rayfin’s migration command to evolve the schema safely without losing data <a class=\"medius-aisummary\" data-starttime=\"00:22:01\">00:22:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:26\">00:24:26</a>. The resulting app supports secure comment creation and deletion rules, reinforcing that schema evolution and fine-grained data security are fundamental advantages of the Rayfin platform. </p>\n\n<p><strong>From Prototype to Production:</strong> After showing the MVP working locally, Lasorsa reveals that production deployment has been silently achieved through the same “Rayfin app” command <a class=\"medius-aisummary\" data-starttime=\"00:28:01\">00:28:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:36\">00:28:36</a>. He opens the live Fabric workspace to show the deployed backend and database and demonstrates that the app now runs publicly and fully managed. He stresses that developers shouldn’t fear the transition from dev to production because Rayfin ensures consistency from day one <a class=\"medius-aisummary\" data-starttime=\"00:29:03\">00:29:03</a>. This integrated approach eliminates the typical “rewrite gap” between prototypes and finished enterprise deployments. Lasorsa then introduces Rayfin’s Analytics App Templates, showing how apps can immediately produce dashboards that visualize live data in Fabric <a class=\"medius-aisummary\" data-starttime=\"00:30:10\">00:30:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:31:56\">00:31:56</a>. </p>\n\n<p><strong>Conclusion and Key Takeaways:</strong> The session concludes by summarizing the core message: the fastest path from idea to production begins with the right foundation <a class=\"medius-aisummary\" data-starttime=\"00:32:27\">00:32:27</a>. Coding agents accelerate development, but sustainable success requires built-in identity, security, governance, and data capabilities. When apps are Fabric-native through platforms like Rayfin, developers gain instant access to analytics and AI-driven insights without friction between systems. Lasorsa closes by encouraging developers to explore the Rayfin documentation for more in-depth learning, reiterating his central mantra: build fast, but build on a foundation that will take you all the way to production <a class=\"medius-aisummary\" data-starttime=\"00:33:14\">00:33:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:38\">00:33:38</a>. </p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8d61253f-f0fa-49ee-9f91-10e852b48f42","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8d61253f-f0fa-49ee-9f91-10e852b48f42","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8d61253f-f0fa-49ee-9f91-10e852b48f42","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774274078845001rWQf_20260603024803947.avif","speakerIds":["1768245294609build26-1769641530377001ZOhB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.636Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774274078845001rWQf_20260603024803947.avif","ogTitle":"Build fast, not fragile with Rayfin and Microsoft Fabric","ogDescription":"Ship new apps in minutes without hand-rolling auth, data, and hosting. In this session, see how RayFin on Microsoft Fabric gives developers and coding agents a code-first backend with typesafe schemas, APIs, functions, storage, and app hosting. We'll walk through going from idea to working app, with Fabric-native data ready for governance, analytics, and AI from day one.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Speaker Digital","SpeakerId":"1768245294609build26-1769641530377001ZOhB","RegistrantKey":"1768245294609build26-1769641530377001ZOhB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/OD810","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"d74a9f37-27ab-4a79-9457-fb4655b283a3","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sam Basu","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776207093064001cZXL","langLocale":"en-US","sessionId":"1776207093064001cZXL","sessionInstanceId":"1776207093064001cZXL","sessionCode":"ODSP929","sessionCodeNormalized":"EN-US-ODSP929","title":"Build modern .NET apps with Uno Platform, AI, and visual tools","sortTitle":"build modern .net apps with uno platform, ai, and visual tools","description":"See how to build modern cross-platform .NET apps using Uno Platform with AI-powered and visual tools. This demo shows how to accelerate feature delivery, improve runtime performance, and achieve consistent rendering with SkiaSharp. Learn patterns for streamlining development across platforms with shared tooling and IDE support.","aiDescription":"<p><strong>Introduction and Overview:</strong> The video opens with Sam Basu welcoming viewers to Microsoft Build 2026 in San Francisco and online (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>). He introduces the session titled “Modern Cross-Platform .NET Development with AI and Visual Tools,” powered by Uno Platform (<a class=\"medius-aisummary\" data-starttime=\"00:00:09\">00:00:09</a>). Sam explains that Uno Platform helps developers take modern .NET applications to every major platform efficiently. It provides a flexible open-source stack that enables building cross-platform apps for mobile (iOS, Android), web (via WebAssembly), desktop (Windows, macOS, Linux), and embedded systems from a single shared C# and XAML codebase (<a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>). He emphasizes productivity features like theming, extensions, and UI toolkits, all designed to streamline the development lifecycle from any OS or IDE such as Visual Studio or VS Code (<a class=\"medius-aisummary\" data-starttime=\"00:01:05\">00:01:05</a>).</p>\n\n<p><strong>Uno Platform Studio and AI Tooling:</strong> Building on this foundation, Sam introduces Uno Platform Studio, a collection of AI-driven design tools crafted to enhance productivity (<a class=\"medius-aisummary\" data-starttime=\"00:01:25\">00:01:25</a>). He demonstrates “Hot Design,” a runtime visual designer that enables developers to tweak their UI while the app is running, keeping everything synchronized through Hot Reload (<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>). He explains the concept of MCP (Model Context Protocol) tools, which bridge the gap between AI agents and real application context by grounding them in documentation and allowing them to validate generated UIs (<a class=\"medius-aisummary\" data-starttime=\"00:02:02\">00:02:02</a>). To show enterprise-grade success, Sam highlights Kahua, a partner using Uno Platform’s AI and design tools to modernize .NET applications and deploy them cross-platform with AI integrated into their user experience (<a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>).</p>\n\n<p><strong>Getting Started and Visual Studio Demo:</strong> The next section demonstrates how to begin with Uno Platform through the official website and documentation (<a class=\"medius-aisummary\" data-starttime=\"00:02:54\">00:02:54</a>). Sam walks through setting up a new cross-platform .NET app in Visual Studio using the Uno Platform extension and the “Uno Check” tool for installing dependencies (<a class=\"medius-aisummary\" data-starttime=\"00:03:34\">00:03:34</a>). He shows the wizard for creating new apps, which allows configuring runtime targets, themes, features, and CI/CD options (<a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>). The scaffolded project supports multiple platforms, maintaining a single shared codebase (<a class=\"medius-aisummary\" data-starttime=\"00:04:26\">00:04:26</a>). Running the default project demonstrates Hot Reload and Hot Design pre-integrated, confirming that Visual Studio-based development is ready for instant iteration (<a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>).</p>\n\n<p><strong>Advanced App Design in Visual Studio Code:</strong> Sam transitions to a Mac setup using Visual Studio Code to show “Uno Chefs,” a more complex sample app from Uno’s open-source gallery (<a class=\"medius-aisummary\" data-starttime=\"00:05:59\">00:05:59</a>). He demonstrates Hot Design in action, opening the live design canvas to modify running app elements directly (<a class=\"medius-aisummary\" data-starttime=\"00:06:49\">00:06:49</a>). Through “Interactive Mode,” developers can navigate their app as users would and modify UI components, such as changing a chart’s color dynamically to verify real-time updates (<a class=\"medius-aisummary\" data-starttime=\"00:08:01\">00:08:01</a>). Sam highlights that Hot Design and Hot Reload work seamlessly together to synchronize UI changes between IDE code, the running app, and the visual design canvas (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>).</p>\n\n<p><strong>AI-Driven App Building and MCP Integration:</strong> Shifting focus to AI-powered workflows, Sam introduces how AI agents like GitHub Copilot, Claude Code, or Gemini can generate and validate .NET UIs using Uno Platform with MCP integration (<a class=\"medius-aisummary\" data-starttime=\"00:09:56\">00:09:56</a>). He explains two MCP servers: Uno MCP, which grounds the AI in Uno documentation, and Uno App MCP, which lets AI interact visually with apps—taking screenshots, navigating UI trees, and simulating user actions (<a class=\"medius-aisummary\" data-starttime=\"00:10:39\">00:10:39</a>). Sam shows how an AI agent can compile, launch an app, and extract data visually, demonstrating a generated enterprise sales dashboard (<a class=\"medius-aisummary\" data-starttime=\"00:12:13\">00:12:13</a>). He also performs doc-grounded AI queries about MVUX design patterns, showing that the agent fetches and summarizes relevant material directly within the IDE (<a class=\"medius-aisummary\" data-starttime=\"00:14:50\">00:14:50</a>). The same experience extends to terminal-based agentic workflows, maintaining consistent contextual understanding (<a class=\"medius-aisummary\" data-starttime=\"00:16:22\">00:16:22</a>).</p>\n\n<p><strong>Showcase, Invitation, and Conclusion:</strong> To illustrate the creative possibilities, Sam highlights Uno’s AI Gallery—a set of visually stunning, AI-generated sample apps like dashboards, animations, and dynamic UI concepts built using the same MCP foundation (<a class=\"medius-aisummary\" data-starttime=\"00:17:18\">00:17:18</a>). He encourages developers to experiment and build their own innovative interfaces using AI and Uno’s design features. In closing, Sam invites attendees to meet the Uno Platform team at Microsoft Build, attend their theater sessions in collaboration with Kahua, and explore upcoming announcements (<a class=\"medius-aisummary\" data-starttime=\"00:18:26\">00:18:26</a>). The video ends with a call to action for developers to begin building at platform.uno and join the community in exploring modern cross-platform .NET development powered by AI and visual design tools (<a class=\"medius-aisummary\" data-starttime=\"00:19:30\">00:19:30</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/007f6aac-9c60-4853-b8e9-b4bfba63ec1d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/007f6aac-9c60-4853-b8e9-b4bfba63ec1d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/007f6aac-9c60-4853-b8e9-b4bfba63ec1d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/007f6aac-9c60-4853-b8e9-b4bfba63ec1d","speakerIds":["1768245294609build26-1776201109444001xetr"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.903Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Uno Platform"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/007f6aac-9c60-4853-b8e9-b4bfba63ec1d","ogTitle":"Build modern .NET apps with Uno Platform, AI, and visual tools","ogDescription":"See how to build modern cross-platform .NET apps using Uno Platform with AI-powered and visual tools. This demo shows how to accelerate feature delivery, improve runtime performance, and achieve consistent rendering with SkiaSharp. Learn patterns for streamlining development across platforms with shared tooling and IDE support.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Speaker Digital","SpeakerId":"1768245294609build26-1776201109444001xetr","RegistrantKey":"1768245294609build26-1776201109444001xetr"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":".NET","logicalValue":".NET"}]},{"@search.score":1.0,"documentId":"bb82bfee-3c5c-43fc-be02-0b534be3ed31","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ravish Patel","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774458839815001F9rf","langLocale":"en-US","sessionId":"1774458839815001F9rf","sessionInstanceId":"1774458839815001F9rf","sessionCode":"ODSP918","sessionCodeNormalized":"EN-US-ODSP918","title":"Build persistent and scalable AI agent memory with TiDB","sortTitle":"build persistent and scalable ai agent memory with tidb","description":"AI agents need different data infrastructure than humans—and they forget everything between sessions. See how TiDB is built for agentic workloads, combining vector search, BM25, and SQL in a unified table to manage agent memory at scale. Learn how hybrid retrieval with RRF, Azure OpenAI embeddings, and ACID transactions enables reliable, scalable agent systems you can deploy in Azure environments.","aiDescription":"<p><strong>Introduction and Problem Context:</strong> The video begins with Ravish Patel introducing himself as a solutions engineer at PingCAP working on TiDB (<a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>). He outlines that the talk will cover building agent memory for AI systems using only SQL, with a demo in the middle (<a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>). To set the foundation, he explains how agent workloads differ from traditional applications. Agents have bursty traffic, massive concurrency, and a continuous need for context recall (<a class=\"medius-aisummary\" data-starttime=\"00:00:28\">00:00:28</a>). These patterns make traditional database designs inefficient and costly when scaling AI agents that need to maintain their own states and memory.</p>\n\n<p><strong>Challenges with Current Multi-Database Architectures:</strong> Patel illustrates how most teams currently solve agent memory by combining three databases—a transactional one for chat history, a vector database for semantic recall, and a search engine for keyword lookups—connected via complex ETL jobs (<a class=\"medius-aisummary\" data-starttime=\"00:01:33\">00:01:33</a>). He details the operational pitfalls of this setup: stale data due to indexing delays, read-after-write inconsistencies (<a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>), partial writes during multi-step operations, and extreme connection fan-out problems as agents scale (<a class=\"medius-aisummary\" data-starttime=\"00:04:32\">00:04:32</a>). These issues often surface only under real production loads, making them hard to catch during early development (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>).</p>\n\n<p><strong>Introducing TiDB and Its Key Capabilities:</strong> To address this, Patel introduces TiDB, a distributed SQL database that scales horizontally with no manual sharding (<a class=\"medius-aisummary\" data-starttime=\"00:05:15\">00:05:15</a>). It is MySQL-compatible, supports transactional, analytical, and vector capabilities within a single system, and is already in production with customers like Pinterest and Dify (<a class=\"medius-aisummary\" data-starttime=\"00:05:58\">00:05:58</a>). TiDB simplifies agent memory architectures by unifying relational data, vector storage, and full-text search in one engine (<a class=\"medius-aisummary\" data-starttime=\"00:06:13\">00:06:13</a>). Features such as agent state tracking, semantic recall with vector indexes, full-text multilingual search, automatic embeddings via SQL functions, and hybrid retrieval provide all the memory tools needed for AI agents without external pipelines (<a class=\"medius-aisummary\" data-starttime=\"00:06:29\">00:06:29</a>).</p>\n\n<p><strong>Performance and Scalability Advantages:</strong> Before the demo, Patel highlights architectural benefits: TiDB supports “scale to zero” for idle agents, so users pay only for active workloads (<a class=\"medius-aisummary\" data-starttime=\"00:07:43\">00:07:43</a>). It can spawn isolated database branches for individual agents in milliseconds and offers resource control to prevent runaway workloads from disrupting others (<a class=\"medius-aisummary\" data-starttime=\"00:08:02\">00:08:02</a>). Its separation of compute and storage allows near-instant scaling during traffic spikes, making it highly adaptable to the unpredictable load patterns of AI systems (<a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>).</p>\n\n<p><strong>Demonstration: Building and Querying Agent Memory:</strong> In the demo (<a class=\"medius-aisummary\" data-starttime=\"00:08:43\">00:08:43</a>), Patel shows the creation of a single TiDB table with user, content, timestamp, and an embedding vector column that automatically generates embeddings using Azure OpenAI upon insert (<a class=\"medius-aisummary\" data-starttime=\"00:09:09\">00:09:09</a>). He inserts five records—covering user data like “Jazz” and “Tokyo Flight”—and demonstrates semantic search (<a class=\"medius-aisummary\" data-starttime=\"00:10:55\">00:10:55</a>), full-text keyword matching (<a class=\"medius-aisummary\" data-starttime=\"00:11:22\">00:11:22</a>), and hybrid search combining both approaches via reciprocal rank fusion (<a class=\"medius-aisummary\" data-starttime=\"00:12:04\">00:12:04</a>). He concludes the technical demo by running an ACID transaction across two tables—memories and user facts—showing consistent updates across both in a single commit (<a class=\"medius-aisummary\" data-starttime=\"00:13:16\">00:13:16</a>).</p>\n\n<p><strong>Customer Success Stories and Closing:</strong> Patel wraps up by sharing real-world results from companies deploying TiDB at scale. Manus runs over 1.5 million agent databases, Dify consolidated half a million containers into one TiDB system, and Pinterest achieved over 1.3 million queries per second using a unified data engine (<a class=\"medius-aisummary\" data-starttime=\"00:14:19\">00:14:19</a>). He recaps key benefits—automatic embeddings, unified query capabilities, hybrid search, resource isolation, and scale-to-zero performance (<a class=\"medius-aisummary\" data-starttime=\"00:15:26\">00:15:26</a>). Finally, he directs viewers to TiDB.com to launch a free cluster in 30 seconds and explore the Python SDK, agentic patterns, and GitHub integrations (<a class=\"medius-aisummary\" data-starttime=\"00:16:04\">00:16:04</a>). The video closes with a call to simplify AI development: “Stop running three databases when you only need one” (<a class=\"medius-aisummary\" data-starttime=\"00:17:01\">00:17:01</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/2ba4cfc6-63e4-4c2c-bedc-0884e1141a16","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/2ba4cfc6-63e4-4c2c-bedc-0884e1141a16","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/2ba4cfc6-63e4-4c2c-bedc-0884e1141a16","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/2ba4cfc6-63e4-4c2c-bedc-0884e1141a16","speakerIds":["1768245294609build26-1778870900847001BXmN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.641Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["TiDB powered by PingCAP"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/2ba4cfc6-63e4-4c2c-bedc-0884e1141a16","ogTitle":"Build persistent and scalable AI agent memory with TiDB","ogDescription":"AI agents need different data infrastructure than humans—and they forget everything between sessions. See how TiDB is built for agentic workloads, combining vector search, BM25, and SQL in a unified table to manage agent memory at scale. Learn how hybrid retrieval with RRF, Azure OpenAI embeddings, and ACID transactions enables reliable, scalable agent systems you can deploy in Azure environments.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778870900847001BXmN","RegistrantKey":"1768245294609build26-1778870900847001BXmN"}],"repeatedSessions":[],"relatedResources":[{"link":"https://static.rainfocus.com/microsoft/build26/sess/1774458839815001F9rf/supmat/ODSP918_PingCAP_Slides_V1_Build26_1779317787323001WJzX.pdf","backpackLink":null,"label":"","category":"Download"}],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Storage","logicalValue":"Storage"},{"displayValue":"Vector Embeddings","logicalValue":"Vector Embeddings"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"8aaa4a0a-8813-401e-97f1-023302c812a1","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jesse Hall","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777652128448001XTsC","langLocale":"en-US","sessionId":"1777652128448001XTsC","sessionInstanceId":"1777652128448001XTsC","sessionCode":"ODSP937","sessionCodeNormalized":"EN-US-ODSP937","title":"Build realtime multimodal agents with LiveKit and Azure","sortTitle":"build realtime multimodal agents with livekit and azure","description":"Multimodal agents demand low latency, network resilience, clean audio, and turn detection all at once. In this session Jesse Hall (Developer Advocate, LiveKit) walks through building a real-time voice agent end-to-end, using Azure's STT, LLM, and TTS models with LiveKit's real-time infrastructure. Get an in-depth look into how the stack fits together and watch the completed agent answer questions about Microsoft Build.","aiDescription":"<p><strong>Introduction to Real-Time Multimodal Agents:</strong> The presentation opens with Jesse Hall introducing the concept of <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:05\">00:00:05</a> real-time multimodal agents, distinguishing them from the single-modality text agents common today. He emphasizes that while latency in text-based systems is tolerable, it becomes critical in real-time interactions. The session’s focus is announced — exploring LiveKit, the cascaded pipeline, and how Azure integrates with LiveKit <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>.</p>\n\n<p><strong>Challenges in Building Real-Time Agentic Systems:</strong> Jesse outlines several technical hurdles involved in constructing real-time multimodal agents <a class=\"medius-aisummary\" data-starttime=\"00:00:42\">00:00:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>. Although models have become easier to deploy, achieving low latency over uncontrolled networks is difficult. He details the steps of capturing the user’s voice, transmitting it, generating a reply, and returning audio within milliseconds. Considerations such as network conditions, voice activity and turn detection, interrupt management, and noise cancellation must all be optimized. Additionally, scalability is emphasized — the system should support potentially hundreds of thousands of concurrent users <a class=\"medius-aisummary\" data-starttime=\"00:01:52\">00:01:52</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>. Jesse stresses the importance of robust real-time media infrastructure and introduces WebRTC as a standard enabling low-latency communication <a class=\"medius-aisummary\" data-starttime=\"00:02:13\">00:02:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:21\">00:02:21</a>.</p>\n\n<p><strong>Introducing LiveKit and Its Architecture:</strong> The speaker explains how LiveKit simplifies the complexity of real-time communication by providing an open-source media layer built on WebRTC <a class=\"medius-aisummary\" data-starttime=\"00:02:24\">00:02:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:31\">00:02:31</a>. It supports not only voice but also video and data transfer, with SDKs available in Python and TypeScript, enabling agents to function across diverse devices — from phones and laptops to embedded systems and robotics <a class=\"medius-aisummary\" data-starttime=\"00:02:38\">00:02:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:55\">00:02:55</a>. He notes that LiveKit is the transport layer utilized by ChatGPT’s voice mode <a class=\"medius-aisummary\" data-starttime=\"00:03:02\">00:03:02</a>. The foundational concept of the voice or “cascaded” pipeline is then introduced: speech passes through voice activity detection (VAD), speech-to-text, large language model processing, and text-to-speech conversion before returning to the user <a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:42\">00:03:42</a>.</p>\n\n<p><strong>LiveKit Infrastructure and Model Agnosticism:</strong> Jesse provides a detailed walkthrough of the LiveKit architecture, showing how the client connects via WebRTC to the LiveKit server that handles real-time transport <a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:11\">00:04:11</a>. Above this is the application layer containing the agent’s SDK, which links media to the models and custom business logic <a class=\"medius-aisummary\" data-starttime=\"00:04:20\">00:04:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:42\">00:04:42</a>. LiveKit itself is model agnostic, allowing interchangeable use of various components for speech recognition (STT), large language processing (LLM), and voice synthesis (TTS) <a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:26\">00:05:26</a>. This flexibility makes integration with Azure straightforward — developers can choose Azure’s STT, OpenAI through Azure for the LLM, and Azure’s TTS <a class=\"medius-aisummary\" data-starttime=\"00:05:30\">00:05:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:41\">00:05:41</a>.</p>\n\n<p><strong>Development Workflow and Best Practices:</strong> Jesse emphasizes leveraging LiveKit’s MCP documentation server to access current and accurate resources for agent development <a class=\"medius-aisummary\" data-starttime=\"00:05:47\">00:05:47</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:16\">00:06:16</a>. He explains that new updates are shipped weekly, so staying synced with documentation is crucial. Moving into a live demo <a class=\"medius-aisummary\" data-starttime=\"00:06:17\">00:06:17</a>, he demonstrates how to create a new agent through the terminal using the AgentStarterPython template and a ReactStarter template for the front-end <a class=\"medius-aisummary\" data-starttime=\"00:06:28\">00:06:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:38\">00:06:38</a>. The setup involves creating the Agent.py file, importing LiveKit and Azure plugins, defining assistant behavior, greetings, VAD, and assigning session identifiers for front-end connectivity <a class=\"medius-aisummary\" data-starttime=\"00:06:41\">00:06:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:56\">00:07:56</a>. Necessary environment variables from LiveKit Cloud and Azure are added to configure models, after which the agent is executed using simple commands <a class=\"medius-aisummary\" data-starttime=\"00:08:26\">00:08:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:01\">00:09:01</a>.</p>\n\n<p><strong>Demo Interaction and Conclusion:</strong> The presentation concludes with a live voice interaction <a class=\"medius-aisummary\" data-starttime=\"00:09:10\">00:09:10</a> between Jesse and the AI assistant. The synthesized computer voice greets attendees at Microsoft Build and lists technologies it can discuss, including Azure, Windows, Microsoft 365, GitHub, and AI projects <a class=\"medius-aisummary\" data-starttime=\"00:09:11\">00:09:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:23\">00:09:23</a>. When Jesse asks about event dates, the agent accurately replies that Microsoft Build 2026 will be held June 2–3 in San Francisco, also available online <a class=\"medius-aisummary\" data-starttime=\"00:09:32\">00:09:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:46\">00:09:46</a>. Jesse wraps up by encouraging experimentation with integrating Azure and LiveKit, highlighting how easily this real-time multimodal communication setup can be implemented and scaled <a class=\"medius-aisummary\" data-starttime=\"00:09:49\">00:09:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:54\">00:09:54</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/08898b31-c2d7-4d94-84c5-14f9e0e70cef","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/08898b31-c2d7-4d94-84c5-14f9e0e70cef","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/08898b31-c2d7-4d94-84c5-14f9e0e70cef","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/08898b31-c2d7-4d94-84c5-14f9e0e70cef","speakerIds":["1768245294609build26-1779899651709001evC9"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.767Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["LiveKit"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/08898b31-c2d7-4d94-84c5-14f9e0e70cef","ogTitle":"Build realtime multimodal agents with LiveKit and Azure","ogDescription":"Multimodal agents demand low latency, network resilience, clean audio, and turn detection all at once. In this session Jesse Hall (Developer Advocate, LiveKit) walks through building a real-time voice agent end-to-end, using Azure's STT, LLM, and TTS models with LiveKit's real-time infrastructure. Get an in-depth look into how the stack fits together and watch the completed agent answer questions about Microsoft Build.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Session Owner, Speaker Digital","SpeakerId":"1768245294609build26-1779899651709001evC9","RegistrantKey":"1768245294609build26-1779899651709001evC9"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"a82386ca-ee6d-4a8c-a301-c19f04f520c2","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Poorvi Narang, Flora Taagen","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799391349001DvGW","langLocale":"en-US","sessionId":"1776799391349001DvGW","sessionInstanceId":"1776799391349001DvGW","sessionCode":"OD827","sessionCodeNormalized":"EN-US-OD827","title":"Build, deploy, and run Linux workloads on Azure","sortTitle":"build, deploy, and run linux workloads on azure","description":"Explore how Azure empowers Linux and open‑source developers with streamlined, secure by default, cloud‑native services designed for fast startup, reduced operational overhead, and consistency across VMs and container workloads. Discover the latest Azure Linux capabilities. See live deployments, security defaults in action, and consistent workload execution from development through production across Azure VMs and AKS.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/e1301455-23a8-4456-a4d0-12dcce7398d1","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/e1301455-23a8-4456-a4d0-12dcce7398d1","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/e1301455-23a8-4456-a4d0-12dcce7398d1","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/e1301455-23a8-4456-a4d0-12dcce7398d1","speakerIds":["1768245294609build26-1775680212026001g2dP","1768245294609build26-1779138478958001s4Au"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.385Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/e1301455-23a8-4456-a4d0-12dcce7398d1","ogTitle":"Build, deploy, and run Linux workloads on Azure","ogDescription":"Explore how Azure empowers Linux and open‑source developers with streamlined, secure by default, cloud‑native services designed for fast startup, reduced operational overhead, and consistency across VMs and container workloads. Discover the latest Azure Linux capabilities. See live deployments, security defaults in action, and consistent workload execution from development through production across Azure VMs and AKS.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1775680212026001g2dP","RegistrantKey":"1768245294609build26-1775680212026001g2dP"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1779138478958001s4Au","RegistrantKey":"1768245294609build26-1779138478958001s4Au"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"fafeb6bf-ebbd-4295-9346-682dac0c060d","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799279520001qMUn","langLocale":"en-US","sessionId":"1776799279520001qMUn","sessionInstanceId":"1776799279520001qMUn","sessionCode":"OD821","sessionCodeNormalized":"EN-US-OD821","title":"Building Azure DocumentDB on Open-Source Foundations","sortTitle":"building azure documentdb on open-source foundations","description":"How do you turn a high-performance open-source database engine into a globally distributed, fully managed cloud service? In this session, we go under the hood of Azure DocumentDB (prev. Azure Cosmos DB for MongoDB vCore) and explore how it is powered by the open-source DocumentDB engine. You'll learn how Microsoft built a multi-tenant cloud architecture around this core, including service orchestration, elastic scaling, workload isolation, and API compatibility.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/53a4c270-d41b-4a8f-b8a5-224f64d1bf02","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/53a4c270-d41b-4a8f-b8a5-224f64d1bf02","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/53a4c270-d41b-4a8f-b8a5-224f64d1bf02","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/53a4c270-d41b-4a8f-b8a5-224f64d1bf02","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.383Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/53a4c270-d41b-4a8f-b8a5-224f64d1bf02","ogTitle":"Building Azure DocumentDB on Open-Source Foundations","ogDescription":"How do you turn a high-performance open-source database engine into a globally distributed, fully managed cloud service? In this session, we go under the hood of Azure DocumentDB (prev. Azure Cosmos DB for MongoDB vCore) and explore how it is powered by the open-source DocumentDB engine. You'll learn how Microsoft built a multi-tenant cloud architecture around this core, including service orchestration, elastic scaling, workload isolation, and API compatibility.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CosmosDB","logicalValue":"CosmosDB"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"}]},{"@search.score":1.0,"documentId":"9f40dfcc-2b68-4e58-b1e2-7649f54e9d40","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Roth","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779999231161001fgvF","langLocale":"en-US","sessionId":"1779999231161001fgvF","sessionInstanceId":"1779999231161001fgvF","sessionCode":"OD802","sessionCodeNormalized":"EN-US-OD802","title":"Building for the agentic web with .NET 11","sortTitle":"building for the agentic web with .net 11","description":"The demands on modern web apps are increasing. Users expect more performance, airtight security, and even agentic capabilities. What does the next generation of web apps look like on .NET? In .NET 11, ASP.NET Core and Blazor are getting faster and more secure at the core, closely integrated with Aspire for distributed app development, and a new set of building blocks — agents, tools, skills, and components — for building agentic web apps. Get ready to build for the modern agentic web!","aiDescription":"<p><strong>Introduction and Vision for .NET 11:</strong> Daniel Roth opens the session at <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a> by introducing himself as the Principal Product Manager for ASP.NET Core and Blazor. He welcomes viewers to \"Building for the Agentic Web with .NET 11\" and explains how modern web applications continuously raise the bar for performance, security, and observability. Users expect apps that are cloud-native, distributed, interactive, and increasingly \"agentic\"—applications that can reason, use tools, and collaborate with people in real time. Developers are looking to build faster by partnering with AI, and .NET 11 investments address the needs of these demanding web apps. Roth mentions that the roadmap for ASP.NET Core and .NET 11 is available on GitHub, structured around six core themes: addressing top feedback, strengthening foundations, investing in the modern stack, simplifying distributed app development, enabling agentic web apps, and improving AI-assisted development.</p>\n\n<p><strong>Strengthening the Foundation:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:02:51\">00:02:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:48\">00:04:48</a>, Roth dives into the first theme—enhancing fundamentals like performance, security, and observability. He highlights improvements such as reduced TLS handshake overhead, optimized handling of malformed requests, and new Zstandard compression for faster connections. Security enhancements include hardening Kestrel (the cross-platform web server) using AI-based scanning, modernized cross-site request forgery protection leveraging Fetch Metadata headers, and automatic authentication token refresh for SignalR and Blazor Server. Observability gains include native OpenTelemetry semantic convention tags emitted automatically for every request, plus upcoming full telemetry support for Blazor WebAssembly components. Roth emphasizes that upgrading to .NET 11 brings faster, safer, and more observable apps without code changes.</p>\n\n<p><strong>Modern Stack Innovations and Demo:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:04:52\">00:04:52</a> and extending through <a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>, Roth details investments into the modern ASP.NET Core stack, primarily Minimal APIs, SignalR, and Blazor. Minimal APIs receive async validation support and OpenAPI 3.2 compatibility, while Blazor’s server-side rendering gets major upgrades for static SSR parity with MVC. In several demos, he showcases new Blazor components like Environment Boundary, Label, and DisplayName, plus improvements in QuickGrid event handling and TempData support for cross-navigation persistence. He also illustrates updates in client-side validation, localization, and asynchronous validation workflows. Roth previews runtime consolidation from Mono to CoreCLR, 64-bit memory capabilities, multithreading, Blazor Web Workers template for background tasks, and support for C# unions—all contributing to a richer, scalable, and more expressive web development platform.</p>\n\n<p><strong>Distributed App Development with Aspire Integration:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:21:21\">00:21:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:28:28\">00:28:28</a>, Roth turns to distributed apps and cloud-readiness, focusing on integration with the Aspire ecosystem. Aspire provides orchestration, observability, and configuration management for multi-service architectures. .NET 11 deepens this relationship by enabling Blazor WebAssembly apps to work seamlessly with Aspire, including service discovery, OpenTelemetry tracing, and debug support. Roth demonstrates a Blazor WebAssembly app hosted via Aspire’s new Blazor Gateway service, which eliminates the need for CORS while handling routing, configuration, and metrics collection. This gateway also replaces the legacy Blazor dev server, offering a production-grade hosting experience for distributed applications with improved scalability and full end-to-end telemetry visibility.</p>\n\n<p><strong>Empowering Agentic and AI-driven Web Apps:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:28:30\">00:28:30</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:03\">00:36:03</a>, Roth introduces the “agentic web”—web applications infused with intelligent agents capable of reasoning, planning, and interacting with users. Collaborating with the Microsoft Agent Framework team, .NET 11 supports protocols such as OpenAI Responses, A2A for multi-agent interaction, and AGUI (Agent-User Interaction) for real-time conversational interfaces. He demonstrates Blazor integration with AGUI, showing a Blazor-based version of the AGUI Dojo app where an agent generates responses, executes tools, and interacts with the UI via both backend and front-end tool calls. The demonstration highlights asynchronous streaming, reasoning progress tracking, and human-in-loop workflows for co-creating plans and content with agents. Roth previews future Blazor AI components simplifying agent-based UI experiences.</p>\n\n<p><strong>AI-Assisted Development and Conclusion:</strong> In the final segment starting at <a class=\"medius-aisummary\" data-starttime=\"00:36:12\">00:36:12</a> through <a class=\"medius-aisummary\" data-starttime=\"00:43:36\">00:43:36</a>, Roth discusses improvements to AI-assisted .NET development. The .NET Skills Marketplace and repository supply agent skills for ASP.NET Core and Blazor development, ensuring coding agents understand framework patterns. He shares examples of skills like “Plan UI Change” that teach AI to better structure UI components, resulting in cleaner code. Testing results compare untrained and skill-enhanced AI-generated Blazor Kanban apps, demonstrating significant improvements in componentization and idiomatic code quality. Roth concludes by summarizing .NET 11’s key achievements across performance, modern tooling, distributed app readiness, agentic integrations, and AI empowerment. He invites developers to try .NET 11 previews, explore Aspire and Microsoft AI Extensions, and contribute to building the agentic web of the future.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","speakerIds":["1768245294609build26-1780001082673001qCTA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/193ed653-1a1c-4bb1-b190-d5cdf880fb9a","ogTitle":"Building for the agentic web with .NET 11","ogDescription":"The demands on modern web apps are increasing. Users expect more performance, airtight security, and even agentic capabilities. What does the next generation of web apps look like on .NET? In .NET 11, ASP.NET Core and Blazor are getting faster and more secure at the core, closely integrated with Aspire for distributed app development, and a new set of building blocks — agents, tools, skills, and components — for building agentic web apps. Get ready to build for the modern agentic web!","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"ASP.NET Core","logicalValue":"ASP.NET Core"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1780001082673001qCTA","RegistrantKey":"1768245294609build26-1780001082673001qCTA"}],"repeatedSessions":[],"relatedResources":[{"link":"https://get.dot.net/11","backpackLink":null,"label":"Download .NET 11","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended Next Step","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"1dc9c9da-f8cf-44cf-994e-1a247d1ab634","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Chris Anderson","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774274079632001rcEz","langLocale":"en-US","sessionId":"1774274079632001rcEz","sessionInstanceId":"1774274079632001rcEz","sessionCode":"OD854","sessionCodeNormalized":"EN-US-OD854","title":"Building WinUI Apps with C# First Patterns and AI Assisted Workflows","sortTitle":"building winui apps with c# first patterns and ai assisted workflows","description":"Developers want faster iteration, clearer control flow, and tools that work well with AI assisted coding. WinUI remains the production platform for Windows apps while adapting to these needs. This session shows how emerging C# first patterns improve developer productivity, how experimental work such as Reactor informs future direction, and how these changes fit alongside the XAML based apps developers are shipping today.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/01c11084-8273-4208-a0f3-8a0119789364","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/01c11084-8273-4208-a0f3-8a0119789364","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/01c11084-8273-4208-a0f3-8a0119789364","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/01c11084-8273-4208-a0f3-8a0119789364","speakerIds":["1768245294609build26-1776980346825001e9Nf"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.636Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/01c11084-8273-4208-a0f3-8a0119789364","ogTitle":"Building WinUI Apps with C# First Patterns and AI Assisted Workflows","ogDescription":"Developers want faster iteration, clearer control flow, and tools that work well with AI assisted coding. WinUI remains the production platform for Windows apps while adapting to these needs. This session shows how emerging C# first patterns improve developer productivity, how experimental work such as Reactor informs future direction, and how these changes fit alongside the XAML based apps developers are shipping today.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776980346825001e9Nf","RegistrantKey":"1768245294609build26-1776980346825001e9Nf"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"WinUI","logicalValue":"WinUI"}]},{"@search.score":1.0,"documentId":"9bf907e4-2b13-49c1-a102-8b6cde556b6b","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Sai Tejaswy Mylavarapu, Avinash Nowduru","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774550901427001lrlR","langLocale":"en-US","sessionId":"1774550901427001lrlR","sessionInstanceId":"1774550901427001lrlR","sessionCode":"OD857","sessionCodeNormalized":"EN-US-OD857","title":"Connected Experience APIs for Windows apps","sortTitle":"connected experience apis for windows apps","description":"Users today are on multiple devices, Windows being one of them. As a Windows app developer, use \"connected experience APIs\" to make your app seamlessly discoverable and installable on the shell for user jobs like share, resume and communication. Try them out, not just for higher usage but also drive new installs on Windows, making your app stickier with users across all device ecosystems.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9977894b-c10d-4dc2-9233-dd0d433d057e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9977894b-c10d-4dc2-9233-dd0d433d057e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9977894b-c10d-4dc2-9233-dd0d433d057e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/9977894b-c10d-4dc2-9233-dd0d433d057e","speakerIds":["1768245294609build26-1776178421672001LElC","1768245294609build26-1776178453831001Emoy"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.843Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/9977894b-c10d-4dc2-9233-dd0d433d057e","ogTitle":"Connected Experience APIs for Windows apps","ogDescription":"Users today are on multiple devices, Windows being one of them. As a Windows app developer, use \"connected experience APIs\" to make your app seamlessly discoverable and installable on the shell for user jobs like share, resume and communication. Try them out, not just for higher usage but also drive new installs on Windows, making your app stickier with users across all device ecosystems.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"},{"displayValue":"Java","logicalValue":"Java"},{"displayValue":"C++","logicalValue":"C++"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776178421672001LElC","RegistrantKey":"1768245294609build26-1776178421672001LElC"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776178453831001Emoy","RegistrantKey":"1768245294609build26-1776178453831001Emoy"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows APIs","logicalValue":"Windows APIs"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Windows SDKs","logicalValue":"Windows SDKs"},{"displayValue":"Cross-device","logicalValue":"Cross-device"},{"displayValue":"App Integration","logicalValue":"App Integration"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"}]},{"@search.score":1.0,"documentId":"8c1a0ead-e587-4b45-b1ec-5d2c63933b57","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Joe Keyes, William Sayer","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778187203542001x1G4","langLocale":"en-US","sessionId":"1778187203542001x1G4","sessionInstanceId":"1778187203542001x1G4","sessionCode":"ODSP940","sessionCodeNormalized":"EN-US-ODSP940","title":"Create enterprise AI apps at scale with Bolt and Microsoft","sortTitle":"create enterprise ai apps at scale with bolt and microsoft","description":"AI-powered app creation is changing what teams can build, but speed alone isn’t enough for the enterprise. See how Bolt integrates with Microsoft tools to move from context and conversation to secure, production-ready applications. Learn how to establish shared design systems, define approved assets and environments, and enable teams to iterate and ship features within governance guardrails—accelerating delivery without sacrificing security or code quality.","aiDescription":"<p><strong>Introduction and Vision:</strong> At the start of the video <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>, Will from Bolt introduces the purpose of the session—exploring how Bolt integrates with Microsoft workflows to transform ideas into secure, production-ready applications. He explains that prompt-driven development has empowered creators like designers and product managers to rapidly move from descriptions to functioning prototypes. However, as Will notes <a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:13\">00:01:13</a>, enterprise teams need more than speed: they require governance, integration with established apps, and developer oversight. Bolt's collaboration with Microsoft addresses this, ensuring AI-generated apps fit within existing cloud environments, workflows, and security frameworks.</p>\n\n<p><strong>Enterprise Integration and Feature Overview:</strong> Will elaborates <a class=\"medius-aisummary\" data-starttime=\"00:01:37\">00:01:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:55\">00:01:55</a> on how Bolt operates within trusted Microsoft ecosystems such as Azure and Microsoft 365, enabling teams to create applications consistent with organizational standards. The generated code can be reviewed and extended before deployment. He emphasizes that developers remain in control of templates, environments, and deployment paths while non-developers benefit from Bolt’s rapid creation capabilities. The demo will cover three main components: organizational design systems, the Bolt Command Line Interface (CLI), and the Bolt Interface <a class=\"medius-aisummary\" data-starttime=\"00:02:33\">00:02:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:12\">00:03:12</a>. Each enables different parts of a team to move faster while maintaining governance and consistency.</p>\n\n<p><strong>Copilot and Bolt Collaboration Demo:</strong> The video demonstrates how application creation begins before opening Bolt <a class=\"medius-aisummary\" data-starttime=\"00:03:34\">00:03:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:27\">00:06:27</a>. Starting in Microsoft Copilot, Will shows how a team transforms a conversation or idea for a travel website template into a structured project brief. Copilot generates layout suggestions, SEO elements, and design goals, then passes the prompt directly to Bolt. When Bolt builds the project, it automatically creates a front-end template with hero carousel, SEO optimizations, and image-based layouts. Through interactive tagging of the Bolt Agent, Will demonstrates iterative collaboration, including modifying the layout to emphasize destination photos. This flow—linking Copilot planning and Bolt execution—ensures continuous context without lost requirements or mismatched specifications.</p>\n\n<p><strong>Developer Workflow and Design System Creation:</strong> Joe takes over <a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a> to showcase Bolt from the developer’s perspective. Working in Visual Studio Code, he demonstrates how Bolt CLI allows developers to bundle existing component libraries and create organizational design systems. Using commands in Copilot, Joe builds and publishes a design system within Bolt, enabling teams to reuse approved components when generating new apps. He opens the resulting Storybook instance which visualizes reusable front-end components. Next, he uses his preloaded design system to prompt Bolt to build a travel destinations landing page, showing automatic code generation, dependency bundling through package.json, and potential integration with private NPM registries. The objective, Joe emphasizes, is to let teams build governed applications using shared assets while retaining code quality and oversight.</p>\n\n<p><strong>Collaborative Enhancements and Advanced Functionality:</strong> Will resumes <a class=\"medius-aisummary\" data-starttime=\"00:12:02\">00:12:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:14\">00:14:14</a> to show how Bolt simplifies teamwork and feature expansion. Through shared project URLs and live agent chat, multiple users can comment and queue fixes in real time. To demonstrate, he prompts Bolt to add user authentication and a trip planning modal using prebuilt components. The agent builds a secure email login system synced to a database and modal interface. Extending functionality further, Will creates a server function integrated with the OpenAI GPT‑5 model: when users input travel details and click “Save Trip,” Bolt generates an itinerary dynamically. Secrets and API keys are handled securely inside Bolt’s management settings. The workflow illustrates how both developers and creators can augment applications instantly while maintaining governance and secure processes.</p>\n\n<p><strong>Conclusion and Organizational Impact:</strong> Wrapping up <a class=\"medius-aisummary\" data-starttime=\"00:14:44\">00:14:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:27\">00:15:27</a>, Will publishes the completed site with inbuilt database scanning, ensuring no vulnerabilities before live deployment. He summarizes the session’s lessons: organizations can use Bolt’s design systems, CLI, and interface to build apps more efficiently and securely. With the combined strengths of Microsoft 365 and Azure powering Bolt, enterprises now have a streamlined method to bring contextual data, validated environments, and scalable infrastructure into every application they create. Bolt accelerates ideation while preserving governance and quality, showcasing a unified future for AI-assisted enterprise software development.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/351dfb42-0c17-4134-aedd-d046ce643cc2","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/351dfb42-0c17-4134-aedd-d046ce643cc2","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/351dfb42-0c17-4134-aedd-d046ce643cc2","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/351dfb42-0c17-4134-aedd-d046ce643cc2","speakerIds":["1768245294609build26-1778636254644001mxVO","1768245294609build26-1778636202424001e063"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.772Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["bolt.new"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/351dfb42-0c17-4134-aedd-d046ce643cc2","ogTitle":"Create enterprise AI apps at scale with Bolt and Microsoft","ogDescription":"AI-powered app creation is changing what teams can build, but speed alone isn’t enough for the enterprise. See how Bolt integrates with Microsoft tools to move from context and conversation to secure, production-ready applications. Learn how to establish shared design systems, define approved assets and environments, and enable teams to iterate and ship features within governance guardrails—accelerating delivery without sacrificing security or code quality.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778636254644001mxVO","RegistrantKey":"1768245294609build26-1778636254644001mxVO"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778636202424001e063","RegistrantKey":"1768245294609build26-1778636202424001e063"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"3aeab2e3-0ac4-4653-97c0-26262e2d2854","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jason Beres","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775166491734001IaaK","langLocale":"en-US","sessionId":"1775166491734001IaaK","sessionInstanceId":"1775166491734001IaaK","sessionCode":"ODSP923","sessionCodeNormalized":"EN-US-ODSP923","title":"Create enterprise apps with AI and MCP","sortTitle":"create enterprise apps with ai and mcp","description":"Build a real enterprise app UI using Ignite UI, MCP servers, and Agent Skills. See AI generate data-rich screens, apply design-system themes, add chat-style UI, and accelerate React, Angular, Blazor, Web Components, WinUI, and .NET MAUI work.","aiDescription":"<p><strong>Introduction and Overview:</strong> The presenter, Jason Beres, begins by introducing the topic of creating enterprise applications leveraging AI and low-code capabilities (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:11\">00:00:11</a>). He explains that the focus is on accelerating enterprise app development with minimal coding through Infragistics UI components, which have been refined over 35 years to support desktop, mobile, and web platforms including Blazor, Angular, React, and Web Components (<a class=\"medius-aisummary\" data-starttime=\"00:00:26\">00:00:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>). He also announces the launch of a new WinUI component suite introduced at Build 2026 (<a class=\"medius-aisummary\" data-starttime=\"00:01:55\">00:01:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:03\">00:02:03</a>), highlighting support for MAUI as part of their cross-platform approach.</p>\n\n<p><strong>Exploration of Products and App Builder Overview:</strong> Jason shows how visitors can learn more about Infragistics products such as UI components, Reveal for embedded analytics, and Slingshot for AI-native work management from the company website (<a class=\"medius-aisummary\" data-starttime=\"00:02:34\">00:02:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:52\">00:02:52</a>). He demonstrates Ignite UI’s unified feature set across frameworks and introduces App Builder — a low-code, WYSIWYG environment for creating apps through either drag-and-drop or conversational AI input (<a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:49\">00:03:49</a>). The process involves signing in with Office 365, choosing sample apps like CRM, applying customizable components such as data grids, themes, and layouts, and seeing generated Blazor or React code instantly reflected (<a class=\"medius-aisummary\" data-starttime=\"00:04:07\">00:04:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:21\">00:06:21</a>). He emphasizes the open-source nature of 50 components, enabling developers to begin building freely while also supporting custom theming and routing.</p>\n\n<p><strong>Publishing and Integration with AI Tools:</strong> Moving into a live coding segment, Jason publishes the CRM app from App Builder directly to GitHub, Azure DevOps, or as a downloadable package (<a class=\"medius-aisummary\" data-starttime=\"00:07:14\">00:07:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:41\">00:07:41</a>). He then clones it in Visual Studio Code and begins configuring AI integration via the Ignite UI CLI, installing MCP (Model Context Protocol) servers and task-specific skills for Blazor (<a class=\"medius-aisummary\" data-starttime=\"00:08:37\">00:08:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:44\">00:09:44</a>). These settings enable GitHub Copilot and Claude Code to understand Ignite UI commands, confirming readiness to help generate new app elements automatically (<a class=\"medius-aisummary\" data-starttime=\"00:10:16\">00:10:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:23\">00:10:23</a>). Upon running the project, he verifies that all data grid interactions — filtering, column manipulation, exporting, and pinning — now appear as expected through generated Blazor code without manual scripting (<a class=\"medius-aisummary\" data-starttime=\"00:10:34\">00:10:34</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:53\">00:10:53</a>).</p>\n\n<p><strong>AI-Driven Application Expansion:</strong> Jason leverages agent-based collaboration to extend the app’s functionality by instructing AI to add a new Ignite UI data grid to the Tasks page using cloud-hosted JSON data (<a class=\"medius-aisummary\" data-starttime=\"00:11:03\">00:11:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:05\">00:12:05</a>). The AI automatically binds data, applies a collapsible multi-row header for addresses, adds filtering and export options, and renders images correctly — all done upon command (<a class=\"medius-aisummary\" data-starttime=\"00:12:10\">00:12:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:35\">00:13:35</a>). The result is a pixel-perfect section with fully interactive elements. Building on this, Jason uploads a design screenshot and requests the AI to synthesize a matching application (<a class=\"medius-aisummary\" data-starttime=\"00:13:46\">00:13:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:22\">00:14:22</a>). Within minutes, it generates a functional blue-themed dashboard that closely replicates the supplied image layout, complete with charts and gauges (<a class=\"medius-aisummary\" data-starttime=\"00:15:24\">00:15:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:32\">00:15:32</a>). He underscores that this AI workflow removes design drift and delivers production-ready results using established frameworks and approved design systems.</p>\n\n<p><strong>WinUI Component Launch and Capabilities:</strong> In the final demonstration, Jason highlights their new WinUI product suite released at Build 2026, featuring sophisticated charts, data grids, and the innovative Dashboard Tile component for interactive user customization (<a class=\"medius-aisummary\" data-starttime=\"00:16:23\">00:16:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:11\">00:17:11</a>). He demonstrates switching chart types dynamically and viewing data analysis options, showing consistent API experiences between WinUI, WPF, and web technologies like Blazor and React (<a class=\"medius-aisummary\" data-starttime=\"00:17:23\">00:17:23</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:50\">00:17:50</a>). Through sample XAML code, he notes that even GitHub Copilot can assist in customizing the WinUI interfaces. The video concludes with encouragement to explore Infragistics’ latest offerings — Ignite UI for web, App Builder for AI-aided low-code development, and WinUI components for modern desktop experiences (<a class=\"medius-aisummary\" data-starttime=\"00:18:13\">00:18:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:12\">00:19:12</a>). He closes by thanking viewers and inviting them to visit infragistics.com for demos and resources to begin building smarter enterprise apps today.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ca45af6a-86b5-4082-921a-c2c8f5fe346e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ca45af6a-86b5-4082-921a-c2c8f5fe346e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ca45af6a-86b5-4082-921a-c2c8f5fe346e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ca45af6a-86b5-4082-921a-c2c8f5fe346e","speakerIds":["1768245294609build26-1776881331857001OX5r"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.845Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Infragistics"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ca45af6a-86b5-4082-921a-c2c8f5fe346e","ogTitle":"Create enterprise apps with AI and MCP","ogDescription":"Build a real enterprise app UI using Ignite UI, MCP servers, and Agent Skills. See AI generate data-rich screens, apply design-system themes, add chat-style UI, and accelerate React, Angular, Blazor, Web Components, WinUI, and .NET MAUI work.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Session Owner, Speaker Digital","SpeakerId":"1768245294609build26-1776881331857001OX5r","RegistrantKey":"1768245294609build26-1776881331857001OX5r"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"c900da26-e52e-4737-aadd-725be45dffdb","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775588883311001gKEH","langLocale":"en-US","sessionId":"1775588883311001gKEH","sessionInstanceId":"1775588883311001gKEH","sessionCode":"ODSP925","sessionCodeNormalized":"EN-US-ODSP925","title":"Deliver production-ready AI search on unstructured data with RAG","sortTitle":"deliver production-ready ai search on unstructured data with rag","description":"RAG is easy to prototype but difficult to run in production. This session shows how to move from proof of concept to production-ready AI search on unstructured data. Build a simple RAG pipeline, then explore patterns for scaling with agentic RAG, graph-based retrieval, and entity recognition. Learn how to choose the right approach for performance, relevance, and maintainability from real-world examples.","aiDescription":"<p><strong>Introduction and Context:</strong> The video opens with Ed Charbeneau from Progress Software introducing himself (<a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>) and outlining the session’s focus—an introduction to Retrieval Augmented Generation (RAG). He explains that the session will explore why RAG is important, how it compares to other generation methods, and introduce Progress Agentic RAG. Ed begins by emphasizing the modern need to make sense of vast amounts of data now produced daily, referencing earlier predictions from 2016 and how the amount has grown exponentially by 2026 (<a class=\"medius-aisummary\" data-starttime=\"00:00:38\">00:00:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:11\">00:01:11</a>). He sets the stage for how machine learning and artificial intelligence have become essential tools for interpreting the data we generate.</p>\n\n<p><strong>Understanding Augmented Generation:</strong> Before defining RAG, Ed explains Context Augmented Generation (<a class=\"medius-aisummary\" data-starttime=\"00:01:45\">00:01:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:10\">00:02:10</a>). He describes how additional data is placed within the model’s context window—such as YouTube transcripts via the Gemini button—to enable conversational querying of content. He then introduces \"Cache Augmented Generation\" (<a class=\"medius-aisummary\" data-starttime=\"00:02:31\">00:02:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:57\">00:02:57</a>), where projects like Copilot in Visual Studio Code use vector memory to store data and assist coding tasks through semantic search. These examples lay groundwork for understanding how different augmentation forms expand AI’s capabilities by enriching context before generation.</p>\n\n<p><strong>Core Principles of Retrieval Augmented Generation:</strong> Ed transitions to explaining Retrieval Augmented Generation (<a class=\"medius-aisummary\" data-starttime=\"00:03:21\">00:03:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:59\">00:03:59</a>) as a specialized version of contextual augmentation utilizing a vector database for semantic retrieval. Unlike keyword searches, RAG operates through semantic similarity, enabling AI to find related information conceptually rather than by exact phrasing (<a class=\"medius-aisummary\" data-starttime=\"00:03:42\">00:03:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:02\">00:04:02</a>). It also includes citation capabilities for transparent results tracking, ensuring users can trace answers to original documents and verify authenticity. Ed highlights how data is embedded into vector formats, sometimes chunked into smaller segments for efficient retrieval and storage. This approach supports generative AI search where natural language queries can interpret broad, unstructured datasets—like PDFs, videos, and web pages—similar to relational databases, marking a significant advancement in business data accessibility (<a class=\"medius-aisummary\" data-starttime=\"00:05:05\">00:05:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>).</p>\n\n<p><strong>Architecture and Industry Challenges:</strong> Ed delves into the complexity of end-to-end RAG architectures (<a class=\"medius-aisummary\" data-starttime=\"00:05:27\">00:05:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:08\">00:06:08</a>), noting the necessity of integrating multiple components—UI, embedding models, document providers, and evaluation systems—often from different vendors. He describes challenges enterprises face when maintaining in-house RAG stacks, particularly issues with scalability, cost prediction, and lack of agentic features like re-ranking and metric evaluation (<a class=\"medius-aisummary\" data-starttime=\"00:06:19\">00:06:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:46\">00:06:46</a>). Referencing industry discussions, he explains how businesses are shifting toward solutions that incorporate intelligent agents for data quality management and retrieval optimization.</p>\n\n<p><strong>Introducing Progress Agentic RAG:</strong> Ed presents Progress Agentic RAG (<a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:57\">00:07:57</a>), a RAG-as-a-Service platform built to resolve those enterprise issues. The system supports ingestion of numerous file types—video, audio, chat logs—and automatically manages structured and unstructured data. Embedded agents handle tagging, entity extraction, and embedding generation while hybrid search mechanisms combine keyword, semantic, and graph searches. An evaluation metric named REMi continuously monitors system quality and stability. Administrators can oversee the platform through a user dashboard, deploy HTML widgets for rapid search interface creation, or use SDKs in .NET, JavaScript, and Python, alongside REST APIs, to build customized agents and applications.</p>\n\n<p><strong>Demonstration and Conclusion:</strong> Ed concludes with a hands-on demonstration of integrating Progress Agentic RAG with .NET and Blazor (<a class=\"medius-aisummary\" data-starttime=\"00:08:29\">00:08:29</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:41\">00:13:41</a>). He walks through creating a financial dashboard that ingests PDF statements and visualizes extracted data in charts using the C# SDK and Blazor Server. The demo includes live search queries to showcase retrieval—like fetching Apple’s revenue from 2024—and converting searches into widgets for embedding in web pages. He highlights the flexible customization potential through the chart-augmented answer object, mapping JSON-structured results into interactive UI elements powered by the Telerik Blazor components. Ed closes the session by pointing viewers to progress.com and a QR code for additional resources and the downloadable demo, thanking the audience for learning about Retrieval Augmented Generation (<a class=\"medius-aisummary\" data-starttime=\"00:13:55\">00:13:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:18\">00:14:18</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/41fbe114-a17d-4a0e-b411-d8ef15e25811","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/41fbe114-a17d-4a0e-b411-d8ef15e25811","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/41fbe114-a17d-4a0e-b411-d8ef15e25811","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/41fbe114-a17d-4a0e-b411-d8ef15e25811","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Progress"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/41fbe114-a17d-4a0e-b411-d8ef15e25811","ogTitle":"Deliver production-ready AI search on unstructured data with RAG","ogDescription":"RAG is easy to prototype but difficult to run in production. This session shows how to move from proof of concept to production-ready AI search on unstructured data. Build a simple RAG pipeline, then explore patterns for scaling with agentic RAG, graph-based retrieval, and entity recognition. Learn how to choose the right approach for performance, relevance, and maintainability from real-world examples.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"}]},{"@search.score":1.0,"documentId":"a97b87a4-fb9e-4bac-8bd6-9cd5040910fe","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Natalie Kershaw","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776797874299001DBvi","langLocale":"en-US","sessionId":"1776797874299001DBvi","sessionInstanceId":"1776797874299001DBvi","sessionCode":"OD833","sessionCodeNormalized":"EN-US-OD833","title":"Deploy AI offline: Creating apps with Foundry Local","sortTitle":"deploy ai offline: creating apps with foundry local","description":"Discover how to ship AI-powered applications that run entirely on-device — no API keys, no network connection, no per-token costs. Through a hands-on demo that participants can follow along and develop in parallel, we'll build a multimodal app that runs completely offline, and can be shipped cross platform. You'll learn how to build AI experiences that are fast, private, and free from cloud dependencies — ideal for regulated industries and resource-constrained environments.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1776797874299001DBvi_20260513131438075.avif","speakerIds":["1768245294609build26-1775765605222001Oo6f"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1776797874299001DBvi_20260513131438075.avif","ogTitle":"Deploy AI offline: Creating apps with Foundry Local","ogDescription":"Discover how to ship AI-powered applications that run entirely on-device — no API keys, no network connection, no per-token costs. Through a hands-on demo that participants can follow along and develop in parallel, we'll build a multimodal app that runs completely offline, and can be shipped cross platform. You'll learn how to build AI experiences that are fast, private, and free from cloud dependencies — ideal for regulated industries and resource-constrained environments.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1775765605222001Oo6f","RegistrantKey":"1768245294609build26-1775765605222001Oo6f"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"Cross-device","logicalValue":"Cross-device"}]},{"@search.score":1.0,"documentId":"d99498e2-42be-4ec0-adb1-2f201634d552","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Guust Ysebie","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396814904001n5ex","langLocale":"en-US","sessionId":"1774396814904001n5ex","sessionInstanceId":"1774396814904001n5ex","sessionCode":"ODSP916","sessionCodeNormalized":"EN-US-ODSP916","title":"Design systems for every user, including people and LLMs","sortTitle":"design systems for every user, including people and llms","description":"Accessibility is often viewed as a usability feature or compliance requirement, important but secondary. In AI‑driven systems, that no longer holds. LLMs interpret the content we produce, and their reliability depends on structure, semantics, and clarity. Accessible content is easier for humans and machines to understand, process, and reuse. This session reframes accessibility as a systems‑level engineering concern, showing why it outperforms OCR and leads to safer, more predictable AI outcomes.","aiDescription":"<p><strong>Introduction and Objectives:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>, Guust Ysebie opens the session by introducing himself and outlining the topic — designing systems for all users, including both people and large language models (LLMs). He explains that his work focuses on PDF accessibility within enterprise environments and emphasizes the relevance of accessibility across both human and machine consumers. The guiding principles introduced — perceivable, operable, understandable, and robust — are identified as crucial for making digital documents universally accessible <a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>.</p>\n\n<p><strong>Understanding PDF Structures and Challenges for LLMs:</strong> Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:01:01\">00:01:01</a>, Ysebie explains how PDFs function as physical data formats combining text, images, and other media through drawing instructions. These instructions render content at precise coordinates on a canvas <a class=\"medius-aisummary\" data-starttime=\"00:01:19\">00:01:19</a>. While humans interpret these visuals intuitively, LLMs perceive only pixel data, making text extraction and semantic understanding difficult without additional processing. He contrasts the human ability to intuit reading order and meaning with the LLM’s lack of inherent context, emphasizing how missing semantics complicates efficient LLM training and comprehension <a class=\"medius-aisummary\" data-starttime=\"00:02:23\">00:02:23</a>.</p>\n\n<p><strong>Semantic PDFs and Structured Accessibility:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>, the discussion turns to the solution provided by tagged PDFs. Ysebie explains that while early PDFs lacked semantic structure, modern ones can embed metadata defining document elements such as headers, paragraphs, and tables. Using such tags mirrors HTML-like organization while preserving pixel-perfect rendering, providing clarity for assistive technologies and LLMs alike. The addition of metadata simplifies searching, extraction, and accessibility workflows. He notes that this structured model transforms unstructured content into meaning-rich data layers, enabling smarter applications and analysis tools <a class=\"medius-aisummary\" data-starttime=\"00:04:49\">00:04:49</a>.</p>\n\n<p><strong>Demonstration of OCR and Embedded Text Extraction:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:05:26\">00:05:26</a>, Ysebie presents a live demo comparing two methods of data extraction from PDFs: using Docling (OCR-based) and using iText (tag-based). He describes running Docling to convert a multi-page PDF to Markdown <a class=\"medius-aisummary\" data-starttime=\"00:06:34\">00:06:34</a>, noting its reliance on pixel-level interpretation and extended runtime of about 18 seconds <a class=\"medius-aisummary\" data-starttime=\"00:07:33\">00:07:33</a>. Although the OCR output preserves tables and lists, it also incorrectly extracts artifacts like watermarks, which may pose risks in AI training pipelines. By contrast, iText completes extraction in only 0.075 seconds <a class=\"medius-aisummary\" data-starttime=\"00:08:42\">00:08:42</a>, producing structurally accurate data without unintended elements and showing 200x better performance and reliability.</p>\n\n<p><strong>Results and Comparative Analysis:</strong> In the following segment (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:04\">00:10:04</a>), Ysebie compares the semantic accuracy of both methods. He highlights that Docling’s OCR failed to detect sublists correctly due to missing contextual tags, whereas iText accurately maintained hierarchy and author intent. This emphasizes how loss of metadata leads to compounded errors over time, affecting LLM training and business data extraction. Tagged PDF structures thus enable safer, faster, and semantically sound outputs that reflect true document design intent.</p>\n\n<p><strong>Best Practices and Conclusion:</strong> Closing around <a class=\"medius-aisummary\" data-starttime=\"00:10:17\">00:10:17</a>, Ysebie stresses the importance of designing PDFs for understanding, not just for rendering. He advises enabling embedded metadata using appropriate library options and verifying that tags accurately mirror the author’s intent. Doing so unlocks advantages like smarter extraction, easier searching, and accessible documents <a class=\"medius-aisummary\" data-starttime=\"00:10:52\">00:10:52</a>. He concludes by reinforcing that accessibility is not extra work but a strategic investment that strengthens both user experience and AI data infrastructure, leading to competitive gains in document processing efficiency. He ends with an invitation for further questions and collaboration <a class=\"medius-aisummary\" data-starttime=\"00:11:31\">00:11:31</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8e0f3456-29c1-4abe-9ca9-d888335803dd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8e0f3456-29c1-4abe-9ca9-d888335803dd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8e0f3456-29c1-4abe-9ca9-d888335803dd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8e0f3456-29c1-4abe-9ca9-d888335803dd","speakerIds":["1768245294609build26-1776256419140001nBLs"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.641Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Apryse"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8e0f3456-29c1-4abe-9ca9-d888335803dd","ogTitle":"Design systems for every user, including people and LLMs","ogDescription":"Accessibility is often viewed as a usability feature or compliance requirement, important but secondary. In AI‑driven systems, that no longer holds. LLMs interpret the content we produce, and their reliability depends on structure, semantics, and clarity. Accessible content is easier for humans and machines to understand, process, and reuse. This session reframes accessibility as a systems‑level engineering concern, showing why it outperforms OCR and leads to safer, more predictable AI outcomes.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776256419140001nBLs","RegistrantKey":"1768245294609build26-1776256419140001nBLs"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"7da487d7-4f9f-41df-a1e0-5950441424a8","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799203939001Pinn","langLocale":"en-US","sessionId":"1776799203939001Pinn","sessionInstanceId":"1776799203939001Pinn","sessionCode":"OD820","sessionCodeNormalized":"EN-US-OD820","title":"Designing Reliable Multi‑Agent Apps with Azure Cosmos DB","sortTitle":"designing reliable multi‑agent apps with azure cosmos db","description":"AI agents are easy to prototype and hard to ship. Learn how to build a multi‑agent, multi‑tenant app that stays fast and reliable in production. See how Azure Cosmos DB powers agent memory, retrieval, and coordination using document processing, vector and full‑text search, and semantic reranking to deliver low‑latency responses at scale.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a6797292-827d-4a63-abef-b603809b2ab3","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a6797292-827d-4a63-abef-b603809b2ab3","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a6797292-827d-4a63-abef-b603809b2ab3","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a6797292-827d-4a63-abef-b603809b2ab3","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.383Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a6797292-827d-4a63-abef-b603809b2ab3","ogTitle":"Designing Reliable Multi‑Agent Apps with Azure Cosmos DB","ogDescription":"AI agents are easy to prototype and hard to ship. Learn how to build a multi‑agent, multi‑tenant app that stays fast and reliable in production. See how Azure Cosmos DB powers agent memory, retrieval, and coordination using document processing, vector and full‑text search, and semantic reranking to deliver low‑latency responses at scale.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/OD820","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"CosmosDB","logicalValue":"CosmosDB"},{"displayValue":"Azure Cosmos DB","logicalValue":"Azure Cosmos DB"},{"displayValue":"CP&D","logicalValue":"CP&D"}]},{"@search.score":1.0,"documentId":"8fdb684a-404c-4443-9207-c5c618cea371","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Greg Crist","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775146465985001DajW","langLocale":"en-US","sessionId":"1775146465985001DajW","sessionInstanceId":"1775146465985001DajW","sessionCode":"ODSP922","sessionCodeNormalized":"EN-US-ODSP922","title":"Develop a conversational search experience without rebuilding your app","sortTitle":"develop a conversational search experience without rebuilding your app","description":"In this on‑demand demo, see how to use Elastic Agent Builder to add LLM‑powered conversational search to an existing production system. You’ll walk through integrating a hosted model on Azure OpenAI and wiring retrieval into search workflows, resulting in a production‑ready conversational experience you can extend immediately.","aiDescription":"<p><strong>Introduction and Context:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a> Greg Crist, Cloud Ecosystem Architect at Elastic, opens the video by welcoming viewers and introducing his colleague Jonathan Simon, who will present a demo. He explains the exciting union of two rapidly evolving technologies: the Microsoft AI ecosystem and Elastic’s capability to supply user-specific data to AI systems. Beginning around <a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>, Greg outlines Microsoft’s Azure OpenAI and Foundry IQ offerings and emphasizes how Elastic integrates seamlessly with Azure Marketplace, enabling users to deploy production-grade AI search within the Microsoft ecosystem. He clarifies that combining Elastic’s data context with Microsoft’s model catalog yields more grounded and trustworthy AI results. Finally, he previews that Jonathan’s demo will show how Elastic Agent Builder transforms a standard e-commerce site into a conversational AI experience.</p>\n\n<p><strong>Overview of Elastic Solutions:</strong> Jonathan Simon begins his portion at <a class=\"medius-aisummary\" data-starttime=\"00:02:04\">00:02:04</a> by introducing himself and the structure of his presentation, which includes an overview, demo, and architectural walkthrough. Around <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>, he explains Elastic’s three key solutions: Search, Observability, and Security. Search provides lexical, semantic, and agentic search through Elastic Agent Builder; Observability offers application monitoring using AI to analyze logs; and Security supports cybersecurity management through automated workflows for attack discovery and remediation. By <a class=\"medius-aisummary\" data-starttime=\"00:04:02\">00:04:02</a>, Jonathan demonstrates how users can easily start by deploying Elastic Serverless from the Azure Marketplace with a few clicks. He then officially announces the general availability of Elastic Agent Builder at <a class=\"medius-aisummary\" data-starttime=\"00:04:55\">00:04:55</a>, marking its transition from technical preview to production-ready status.</p>\n\n<p><strong>E-Commerce Demo and AI Search Capabilities:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:05:14\">00:05:14</a>, Jonathan launches into a demo of an e-commerce site for a fictional Wayfinder Supply Company. The site’s products, descriptions, and titles are AI-generated, with Elastic powering its search. Between <a class=\"medius-aisummary\" data-starttime=\"00:05:33\">00:05:33</a> and <a class=\"medius-aisummary\" data-starttime=\"00:06:43\">00:06:43</a>, he showcases browsing and hybrid search functions. When searching for \"jackets\" versus \"coats,\" he demonstrates how lexical searches fail on missing keywords while hybrid search succeeds by using vector-based semantic connections via Elastic Inference Service. The demo then evolves at <a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a> as Jonathan envisions creating a trip planner powered by AI — an agent that can recommend camping gear and plan itineraries — which leads into building such an agent through Elastic Agent Builder.</p>\n\n<p><strong>Building the Trip Planner Agent with Elastic Agent Builder:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:07:17\">00:07:17</a>, Jonathan enters the Agent Builder interface. He illustrates how pre-configured large language models (LLMs), such as Anthropic, can be swapped out or expanded by connecting Azure-hosted models. Between <a class=\"medius-aisummary\" data-starttime=\"00:08:05\">00:08:05</a> and <a class=\"medius-aisummary\" data-starttime=\"00:09:51\">00:09:51</a>, he shows the deployment of a Mistral Large model on Microsoft Foundry, linking it via URI and API key to Elastic. Once validated, this custom connector appears as a selectable LLM inside Agent Builder. At <a class=\"medius-aisummary\" data-starttime=\"00:10:02\">00:10:02</a>, Jonathan creates a workflow for the Trip Planner, copying workflow logic from code, explaining how it handles manual triggers, HTTP posts to MCP servers, retries, and logging. By <a class=\"medius-aisummary\" data-starttime=\"00:12:08\">00:12:08</a>, the workflow successfully retrieves sample user data, confirming operational readiness. He then constructs a tool using this workflow at <a class=\"medius-aisummary\" data-starttime=\"00:13:00\">00:13:00</a> and validates its output with customer profile data before finally building the Trip Planner agent around <a class=\"medius-aisummary\" data-starttime=\"00:15:02\">00:15:02</a>, assigning it tools and behavior parameters.</p>\n\n<p><strong>Testing and Integrating the Agent:</strong> With the Trip Planner created, Jonathan tests it inside Agent Builder at <a class=\"medius-aisummary\" data-starttime=\"00:16:19\">00:16:19</a>. Running the prompt “Plan a three-day backpacking trip to Yosemite this weekend,” the agent generates detailed reasoning traces showing weather checks, gear searches, itinerary creation, and trip recommendations. By <a class=\"medius-aisummary\" data-starttime=\"00:18:11\">00:18:11</a>, the finalized itinerary is displayed within Agent Builder. He demonstrates that all these agent operations are API-accessible; at <a class=\"medius-aisummary\" data-starttime=\"00:19:13\">00:19:13</a>, he queries the agent to generate an example Python API call, showing how external apps can use the same functions. Finally, Jonathan integrates the Trip Planner back into the Wayfinder web app at <a class=\"medius-aisummary\" data-starttime=\"00:20:10\">00:20:10</a>, confirming that the button now triggers live trip planning using the AI agent. The app surfaces reasoning steps, recommended gear, itineraries, and direct add-to-cart features, converting simple search into a conversation-driven recommendation engine.</p>\n\n<p><strong>Architecture and Conclusion:</strong> Near <a class=\"medius-aisummary\" data-starttime=\"00:22:42\">00:22:42</a>, Jonathan concludes by exploring the architecture underpinning the demo. The React-based front end sends requests to a FastAPI Python backend, which communicates with Elastic’s Trip Planner agent. The agent accesses four custom tools: semantic product search, user affinity via ESQL queries, trip safety via MCP workflows, and customer profile retrieval through CRM integration. By <a class=\"medius-aisummary\" data-starttime=\"00:24:00\">00:24:00</a>, he encourages viewers to explore Elastic Agent Builder through official documentation, GitHub code, and a free workshop, highlighting that Azure users can deploy a free seven-day trial through a QR code. The video wraps up at <a class=\"medius-aisummary\" data-starttime=\"00:25:02\">00:25:02</a> as Greg Crist thanks Jonathan and invites attendees to visit Elastic’s booth at the Microsoft Build Conference 2026, closing on music and a reminder of the synergy between Elastic and Microsoft AI solutions.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/107cd1bc-d7fd-428c-8a8c-fa58870c3a9a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/107cd1bc-d7fd-428c-8a8c-fa58870c3a9a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/107cd1bc-d7fd-428c-8a8c-fa58870c3a9a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/107cd1bc-d7fd-428c-8a8c-fa58870c3a9a","speakerIds":["1768245294609build26-1764099185727001ifAC"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.845Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Elastic"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/107cd1bc-d7fd-428c-8a8c-fa58870c3a9a","ogTitle":"Develop a conversational search experience without rebuilding your app","ogDescription":"In this on‑demand demo, see how to use Elastic Agent Builder to add LLM‑powered conversational search to an existing production system. You’ll walk through integrating a hosted model on Azure OpenAI and wiring retrieval into search workflows, resulting in a production‑ready conversational experience you can extend immediately.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1764099185727001ifAC","RegistrantKey":"1768245294609build26-1764099185727001ifAC"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Foundry IQ","logicalValue":"Foundry IQ"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Entra","logicalValue":"Entra"}]},{"@search.score":1.0,"documentId":"91421e47-0449-4572-bbc0-7c8f1baf4bec","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kerim Satirli","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774395793920001zqvL","langLocale":"en-US","sessionId":"1774395793920001zqvL","sessionInstanceId":"1774395793920001zqvL","sessionCode":"ODSP901","sessionCodeNormalized":"EN-US-ODSP901","title":"Discover, govern, and scale Azure resources with HashiCorp Terraform","sortTitle":"discover, govern, and scale azure resources with hashicorp terraform","description":"Managing existing Azure environments is a major barrier to scaling cloud operations, with many resources outside infrastructure-as-code. In this session, learn how HashiCorp Terraform and Terraform Search help teams discover unmanaged Azure resources and bring them under management quickly and declaratively. Move from fragmented environments to consistent, policy-driven infrastructure for AI workloads and agents while improving governance, reducing manual effort, and accelerating Azure adoption.","aiDescription":"<p><strong>Introduction and Problem Context:</strong> The session opens with Kerim Satirli introducing himself as a Senior Developer Advocate at HashiCorp and presenting the focus of the talk: discovering, governing, and scaling Microsoft Azure resources using HashiCorp Terraform <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>. He identifies a common challenge faced by Azure teams—the disconnect between Terraform’s state and the actual cloud infrastructure running within subscriptions and tenants. Many assume near-total management through Terraform, yet real audits show only about 40–60% coverage. This mismatch occurs because infrastructure often drifts outside code management for human reasons. Kerim promises that by the end of the next twenty minutes, viewers will see a workflow to close this gap dramatically faster than traditional methods <a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:34\">00:01:34</a>.</p>\n\n<p><strong>Explaining Infrastructure as Code and Causes of Drift:</strong> Kerim then explains \"infrastructure as code\" for those newer to Terraform <a class=\"medius-aisummary\" data-starttime=\"00:01:40\">00:01:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>. Unlike manual Azure portal provisioning, defining infrastructure in code improves repeatability and lowers human error. However, unmanaged resources arise for various human reasons: postponed codification (“I’ll codify later”), emergency fixes <a class=\"medius-aisummary\" data-starttime=\"00:03:01\">00:03:01</a>, acquisitions where environments are inherited <a class=\"medius-aisummary\" data-starttime=\"00:03:17\">00:03:17</a>, and proofs-of-concept that became production without proper rebuild <a class=\"medius-aisummary\" data-starttime=\"00:03:33\">00:03:33</a>. He stresses that these situations are unavoidable, framing the goal not as prevention but as having a fast path to bring unmanaged resources back under control. Kerim also notes that unmanaged infrastructure is particularly problematic for AI workloads—causing misattributed costs, insecure networks, and identity sprawl for agent-based architectures <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:49\">00:04:49</a>.</p>\n\n<p><strong>Introducing Terraform Import and Terraform Search:</strong> Shifting focus to the solution, Kerim introduces how Terraform traditionally handled imports—using CLI commands—with one resource managed at a time, which was slow and manual <a class=\"medius-aisummary\" data-starttime=\"00:05:01\">00:05:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:19\">00:05:19</a>. Later releases introduced import blocks, improving efficiency but still requiring teams to manually specify resource IDs. The recent Terraform 1.14 upgrade, released November 2025, brought a breakthrough feature—Terraform Search <a class=\"medius-aisummary\" data-starttime=\"00:06:09\">00:06:09</a>. This tool automates resource discovery and code generation in three steps: define a query, execute it via CLI, and generate code. With Terraform Search, users can request unmanaged Azure resources (e.g., all SQL servers) and automatically receive configuration files, dramatically reducing manual effort and potential errors <a class=\"medius-aisummary\" data-starttime=\"00:06:24\">00:06:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:10\">00:07:10</a>. The feature inverts the traditional import logic—shifting from “I know what I want” to “show me what’s there and help me manage it”—saving teams weeks of inventory work <a class=\"medius-aisummary\" data-starttime=\"00:07:26\">00:07:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:46\">00:07:46</a>.</p>\n\n<p><strong>Live Demo: Discovering and Importing Unmanaged Azure Resources:</strong> Kerim transitions to a live demo showcasing Terraform Search in action <a class=\"medius-aisummary\" data-starttime=\"00:07:53\">00:07:53</a>. He displays an unmanaged Azure environment containing roughly thirty scattered resources—networking, storage, gateways, and DNS entries—with missing or inconsistent tagging <a class=\"medius-aisummary\" data-starttime=\"00:08:04\">00:08:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:37\">00:08:37</a>. Through a discovery query targeting resources such as NAT gateways, security groups, and public IPs, Kerim demonstrates how Terraform inherits Azure authentication details automatically and lists unmanaged resources for import <a class=\"medius-aisummary\" data-starttime=\"00:10:08\">00:10:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:22\">00:10:22</a>. Exporting the configuration creates a Terraform file containing generated resource definitions and import blocks, which can be applied through a normal Terraform plan/apply cycle <a class=\"medius-aisummary\" data-starttime=\"00:11:42\">00:11:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:56\">00:12:56</a>. This output automates what previously required hundreds of manual lines of code and allows immediate alignment with the Terraform state.</p>\n\n<p><strong>Applying Organizational Governance with Sentinel:</strong> Once resources are imported, Kerim explains the next step—governing them using HashiCorp Sentinel <a class=\"medius-aisummary\" data-starttime=\"00:14:03\">00:14:03</a>. Sentinel enforces policies across Terraform-managed environments, ensuring newly imported resources comply with organizational standards such as proper tags, naming conventions, and approved SKUs. He builds and demonstrates a policy that validates ownership metadata, detecting violations like missing or misformatted tags <a class=\"medius-aisummary\" data-starttime=\"00:15:25\">00:15:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:41\">00:15:41</a>. Kerim notes that Sentinel can operate in soft-fail or hard-fail modes, and organizations should transition toward hard enforcement for strong governance <a class=\"medius-aisummary\" data-starttime=\"00:16:37\">00:16:37</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:54\">00:16:54</a>. This approach eliminates configuration drift, ensures consistency, and helps teams maintain a compliant infrastructure environment.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> The session closes with Kerim summarizing key takeaways—teams can now use Terraform Query to identify and import unmanaged Azure resources, and Sentinel to enforce governance policies <a class=\"medius-aisummary\" data-starttime=\"00:17:12\">00:17:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:26\">00:17:26</a>. He provides helpful resources for further learning, including documentation for CLI commands, a detailed blog about Terraform Query, and a Sentinel playground for testing policies <a class=\"medius-aisummary\" data-starttime=\"00:17:49\">00:17:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:03\">00:18:03</a>. Kerim ends by thanking viewers and wishing them a productive experience at Microsoft Build 2026 <a class=\"medius-aisummary\" data-starttime=\"00:18:07\">00:18:07</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/db635682-1544-4b56-873d-1ac9b8e2509b","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/db635682-1544-4b56-873d-1ac9b8e2509b","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/db635682-1544-4b56-873d-1ac9b8e2509b","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/db635682-1544-4b56-873d-1ac9b8e2509b","speakerIds":["1768245294609build26-1775526992416001GCBk"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.638Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["HashiCorp","an IBM company"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/db635682-1544-4b56-873d-1ac9b8e2509b","ogTitle":"Discover, govern, and scale Azure resources with HashiCorp Terraform","ogDescription":"Managing existing Azure environments is a major barrier to scaling cloud operations, with many resources outside infrastructure-as-code. In this session, learn how HashiCorp Terraform and Terraform Search help teams discover unmanaged Azure resources and bring them under management quickly and declaratively. Move from fragmented environments to consistent, policy-driven infrastructure for AI workloads and agents while improving governance, reducing manual effort, and accelerating Azure adoption.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital, Owner","SpeakerId":"1768245294609build26-1775526992416001GCBk","RegistrantKey":"1768245294609build26-1775526992416001GCBk"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"Terraform","logicalValue":"Terraform"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"fc4a2eba-3a92-4078-9700-b79bcabb92ba","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeremiah Follis","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776898113963001DDsb","langLocale":"en-US","sessionId":"1776898113963001DDsb","sessionInstanceId":"1776898113963001DDsb","sessionCode":"OD840","sessionCodeNormalized":"EN-US-OD840","title":"Enable agents for enterprises using Agent 365 SDK","sortTitle":"enable agents for enterprises using agent 365 sdk","description":"As agents move from prototypes into real business workflows, organizations need a way to make custom and third-party agents visible, governable, and secure at enterprise scale. This session shows how Agent 365 SDK helps bring agents into an enterprise-ready control plane so teams can manage identity, observability, governance, and security with confidence.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/dc54c74e-26f4-4838-b660-c88746833c20","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/dc54c74e-26f4-4838-b660-c88746833c20","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/dc54c74e-26f4-4838-b660-c88746833c20","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/dc54c74e-26f4-4838-b660-c88746833c20","speakerIds":["1768245294609build26-1770939979775001ZoOw"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.387Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/dc54c74e-26f4-4838-b660-c88746833c20","ogTitle":"Enable agents for enterprises using Agent 365 SDK","ogDescription":"As agents move from prototypes into real business workflows, organizations need a way to make custom and third-party agents visible, governable, and secure at enterprise scale. This session shows how Agent 365 SDK helps bring agents into an enterprise-ready control plane so teams can manage identity, observability, governance, and security with confidence.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1770939979775001ZoOw","RegistrantKey":"1768245294609build26-1770939979775001ZoOw"}],"repeatedSessions":[],"relatedResources":[{"link":"https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Faka.ms%2FA365SDK-Docs&data=05%7C02%7Cconexp%40microsoft.com%7Cbab46e8fa86d4bb253aa08debd0b49e7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C639156053337866236%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=f4L4KiTuGcV14%2FqXPwQ1zKMlW0I8m6i7VZJOiHqn6iI%3D&reserved=0","backpackLink":null,"label":"SDK Overview","category":"Web link"},{"link":"https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Faka.ms%2FA365SDK-Blog&data=05%7C02%7Cconexp%40microsoft.com%7Cbab46e8fa86d4bb253aa08debd0b49e7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C639156053337882974%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=D3iw5shn3WEHUpGqrTihFBsVPqr97ft%2F%2BF0ri1XBcDY%3D&reserved=0","backpackLink":null,"label":"Blog","category":"Web link"},{"link":"https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Faka.ms%2FA365SDK-Start&data=05%7C02%7Cconexp%40microsoft.com%7Cbab46e8fa86d4bb253aa08debd0b49e7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C639156053337848683%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=ZEL3AspqDd0A09EDrVvxalIAfPaMGx2f6pKxsbIMVwo%3D&reserved=0","backpackLink":null,"label":"Get Started","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Agent 365","logicalValue":"Agent 365"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Enterprise","logicalValue":"Enterprise"},{"displayValue":"Identity","logicalValue":"Identity"}]},{"@search.score":1.0,"documentId":"80dc1037-7d51-4d90-ae2e-85707ec9d7f1","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Maha Bayana, Andrew Leader","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774274079143001rCao","langLocale":"en-US","sessionId":"1774274079143001rCao","sessionInstanceId":"1774274079143001rCao","sessionCode":"OD851","sessionCodeNormalized":"EN-US-OD851","title":"Expand local AI reach with Windows ML","sortTitle":"expand local ai reach with windows ml","description":"Windows ML lets developers build local, AI‑powered apps using custom or open‑source models that run efficiently across GPU, NPU, and CPU on a unified Windows platform. Learn what’s new, including WebNN support for web experiences and improved tooling with AI Toolkit for VS Code to simplify preparing and deploying models and AI workloads with Windows ML.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/db63ff28-f961-48b0-9768-2e330580c301","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/db63ff28-f961-48b0-9768-2e330580c301","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/db63ff28-f961-48b0-9768-2e330580c301","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/db63ff28-f961-48b0-9768-2e330580c301","speakerIds":["1768245294609build26-1776963523419001iUHD","1768245294609build26-1774285002790001pcAb"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.635Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/db63ff28-f961-48b0-9768-2e330580c301","ogTitle":"Expand local AI reach with Windows ML","ogDescription":"Windows ML lets developers build local, AI‑powered apps using custom or open‑source models that run efficiently across GPU, NPU, and CPU on a unified Windows platform. Learn what’s new, including WebNN support for web experiences and improved tooling with AI Toolkit for VS Code to simplify preparing and deploying models and AI workloads with Windows ML.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1774285002790001pcAb","RegistrantKey":"1768245294609build26-1774285002790001pcAb"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776963523419001iUHD","RegistrantKey":"1768245294609build26-1776963523419001iUHD"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/OD851","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"Foundry Local","logicalValue":"Foundry Local"}]},{"@search.score":1.0,"documentId":"632863b8-bcf4-4633-9f8a-d093bc5c3e1d","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Yitzhak Kesselman, Tessa Kloster","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693090099001GrMO","langLocale":"en-US","sessionId":"1775693090099001GrMO","sessionInstanceId":"1775693090099001GrMO","sessionCode":"OD812","sessionCodeNormalized":"EN-US-OD812","title":"Fabric IQ: Bringing enterprise intelligence into the developer workflow","sortTitle":"fabric iq: bringing enterprise intelligence into the developer workflow","description":"Join this session to learn how Microsoft Fabric IQ helps developers build AI-powered data apps and agents on a secure, governed semantic foundation. Fabric IQ reuses existing semantic models, connects to live data, and captures shared business logic to unify data, meaning, and action across systems. See how a shared ontology reduces brittle integrations and gives agents durable business context using MCP, GitHub Copilot, and Ontology APIs and SDKs.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/9cdbf41a-4e85-40ae-9ce3-d4625377ab49","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/9cdbf41a-4e85-40ae-9ce3-d4625377ab49","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/9cdbf41a-4e85-40ae-9ce3-d4625377ab49","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693090099001GrMO_20260603024834209.avif","speakerIds":["1768245294609build26-1765523893273001iiJj","1768245294609build26-1765361984218001ZKhA"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693090099001GrMO_20260603024834209.avif","ogTitle":"Fabric IQ: Bringing enterprise intelligence into the developer workflow","ogDescription":"Join this session to learn how Microsoft Fabric IQ helps developers build AI-powered data apps and agents on a secure, governed semantic foundation. Fabric IQ reuses existing semantic models, connects to live data, and captures shared business logic to unify data, meaning, and action across systems. See how a shared ontology reduces brittle integrations and gives agents durable business context using MCP, GitHub Copilot, and Ontology APIs and SDKs.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) Digital","SpeakerId":"1768245294609build26-1765523893273001iiJj","RegistrantKey":"1768245294609build26-1765523893273001iiJj"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1765361984218001ZKhA","RegistrantKey":"1768245294609build26-1765361984218001ZKhA"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/OD812","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"2e24d662-1e3e-4985-b0a1-4f20a9f5f540","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Philip Laussermair","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799345079001Dc7Z","langLocale":"en-US","sessionId":"1776799345079001Dc7Z","sessionInstanceId":"1776799345079001Dc7Z","sessionCode":"OD823","sessionCodeNormalized":"EN-US-OD823","title":"Faster AI Responses with Semantic Caching in Azure Managed Redis","sortTitle":"faster ai responses with semantic caching in azure managed redis","description":"AI apps are only as efficient as how they handle repeated intent. In this demo, see how semantic caching with Azure Managed Redis identifies similar requests and serves responses instantly—without reprocessing tokens. Compare a standard LLM workflow to a cached experience in real time, and learn how vector similarity search reduces latency, cuts token usage, and lowers cost for scalable AI apps.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/5501ab3e-c66b-45a9-af7f-73a90cb8fc9d","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/5501ab3e-c66b-45a9-af7f-73a90cb8fc9d","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/5501ab3e-c66b-45a9-af7f-73a90cb8fc9d","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/5501ab3e-c66b-45a9-af7f-73a90cb8fc9d","speakerIds":["1768245294609build26-1759173805963001hLfB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/5501ab3e-c66b-45a9-af7f-73a90cb8fc9d","ogTitle":"Faster AI Responses with Semantic Caching in Azure Managed Redis","ogDescription":"AI apps are only as efficient as how they handle repeated intent. In this demo, see how semantic caching with Azure Managed Redis identifies similar requests and serves responses instantly—without reprocessing tokens. Compare a standard LLM workflow to a cached experience in real time, and learn how vector similarity search reduces latency, cuts token usage, and lowers cost for scalable AI apps.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1759173805963001hLfB","RegistrantKey":"1768245294609build26-1759173805963001hLfB"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL","logicalValue":"Azure SQL"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"0cdd98d5-4d6f-4063-b7b0-952f0ac8f5af","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jeff Blankenburg","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396081838001IZ4C","langLocale":"en-US","sessionId":"1774396081838001IZ4C","sessionInstanceId":"1774396081838001IZ4C","sessionCode":"ODSP905","sessionCodeNormalized":"EN-US-ODSP905","title":"Fix your broken debugging workflow","sortTitle":"fix your broken debugging workflow","description":"Code velocity isn’t the problem. Understanding real-time systems behavior is. AI and agents can ship changes fast, but without runtime context they’re guessing.\n\nThe old loop doesn’t scale:\ndeploy → dashboards → logs → redeploy\n\nIn this session, build, break, and fix a real service in VS Code using Copilot agent mode and Dynatrace MCP. See runtime context flow into the editor so debugging becomes continuous understanding and both you and your agents operate from real system behavior.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"","hasOnDemand":false,"downloadVideoLink":"","captionFileLink":"","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1774396081838001IZ4C_20260512023758852.avif","speakerIds":["1768245294609build26-1775666678371001JTOl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.639Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Dynatrace"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1774396081838001IZ4C_20260512023758852.avif","ogTitle":"Fix your broken debugging workflow","ogDescription":"Code velocity isn’t the problem. Understanding real-time systems behavior is. AI and agents can ship changes fast, but without runtime context they’re guessing.\n\nThe old loop doesn’t scale:\ndeploy → dashboards → logs → redeploy\n\nIn this session, build, break, and fix a real service in VS Code using Copilot agent mode and Dynatrace MCP. See runtime context flow into the editor so debugging becomes continuous understanding and both you and your agents operate from real system behavior.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1775666678371001JTOl","RegistrantKey":"1768245294609build26-1775666678371001JTOl"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Outages","logicalValue":"Outages"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Copilot","logicalValue":"Copilot"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Deployment Pipelines","logicalValue":"Deployment Pipelines"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"Visual Studio","logicalValue":"Visual Studio"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"}]},{"@search.score":1.0,"documentId":"96932c90-346e-4b5a-b707-fd972eb63c38","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Carlotta Castelluccio​, Nitya Narasimhan","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777498707394001lLOw","langLocale":"en-US","sessionId":"1777498707394001lLOw","sessionInstanceId":"1777498707394001lLOw","sessionCode":"OD836","sessionCodeNormalized":"EN-US-OD836","title":"From Test Kitchen to Table: A Demo-driven Tour of Foundry Portal for AI Developers","sortTitle":"from test kitchen to table: a demo-driven tour of foundry portal for ai developers","description":"Follow a live journey from rapid experimentation in the Foundry portal's playground to production-ready agent deployment. Discover how the portal's UI streamlines workflow design, model management, and monitoring, making it easy for developers to discover, build, iterate and manage their AI solutions in a manner that complements their code-first mindset. By mapping each portal capability to familiar developer tasks, we'll also reinforce how the portal enhances the code-first experience with richer observability features to give them a more comprehensive understanding of system behaviors that may not be visible through code alone.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/c21b357d-838a-471d-a428-387fcaa2f602","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/c21b357d-838a-471d-a428-387fcaa2f602","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/c21b357d-838a-471d-a428-387fcaa2f602","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/c21b357d-838a-471d-a428-387fcaa2f602","speakerIds":["1768245294609build26-1756142402845001TiE3","1768245294609build26-1756141269010001LleM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.389Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/c21b357d-838a-471d-a428-387fcaa2f602","ogTitle":"From Test Kitchen to Table: A Demo-driven Tour of Foundry Portal for AI Developers","ogDescription":"Follow a live journey from rapid experimentation in the Foundry portal's playground to production-ready agent deployment. Discover how the portal's UI streamlines workflow design, model management, and monitoring, making it easy for developers to discover, build, iterate and manage their AI solutions in a manner that complements their code-first mindset. By mapping each portal capability to familiar developer tasks, we'll also reinforce how the portal enhances the code-first experience with richer observability features to give them a more comprehensive understanding of system behaviors that may not be visible through code alone.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1756142402845001TiE3","RegistrantKey":"1768245294609build26-1756142402845001TiE3"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1756141269010001LleM","RegistrantKey":"1768245294609build26-1756141269010001LleM"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/build/foundrydiscord","backpackLink":null,"label":"Microsoft Foundry in Discord","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26/OD836","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"4d61436d-7bd0-4cf1-be9b-79ac6d0cb6bb","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Divya Swarnkar","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776797629916001PUri","langLocale":"en-US","sessionId":"1776797629916001PUri","sessionInstanceId":"1776797629916001PUri","sessionCode":"OD832","sessionCodeNormalized":"EN-US-OD832","title":"From workflows to agentic automation with Azure Logic Apps","sortTitle":"from workflows to agentic automation with azure logic apps","description":"Learn how to build and operationalize agentic business processes with Azure Logic Apps Automation, a new offering purpose-built for agentic business process automation. Discover how to orchestrate workflows, APIs, Microsoft Foundry agents, and human decisions across enterprise systems. Through demos, see how to design, develop, and operationalize agentic workflows with reliability, observability, and governance built in.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/45a66455-ef27-45ef-af08-16232ea239ad","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/45a66455-ef27-45ef-af08-16232ea239ad","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/45a66455-ef27-45ef-af08-16232ea239ad","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/45a66455-ef27-45ef-af08-16232ea239ad","speakerIds":["1768245294609build26-1777934900615001XSkq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.383Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/45a66455-ef27-45ef-af08-16232ea239ad","ogTitle":"From workflows to agentic automation with Azure Logic Apps","ogDescription":"Learn how to build and operationalize agentic business processes with Azure Logic Apps Automation, a new offering purpose-built for agentic business process automation. Discover how to orchestrate workflows, APIs, Microsoft Foundry agents, and human decisions across enterprise systems. Through demos, see how to design, develop, and operationalize agentic workflows with reliability, observability, and governance built in.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777934900615001XSkq","RegistrantKey":"1768245294609build26-1777934900615001XSkq"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Development pipeline","logicalValue":"Development pipeline"}]},{"@search.score":1.0,"documentId":"c87c5b96-d4f7-4e27-9f01-f88ecb6b406b","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Mike Budzynski, Anish Tallapureddy","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776797688849001ZK6A","langLocale":"en-US","sessionId":"1776797688849001ZK6A","sessionInstanceId":"1776797688849001ZK6A","sessionCode":"OD831","sessionCodeNormalized":"EN-US-OD831","title":"Govern AI models, tools, and agents with Azure API Management","sortTitle":"govern ai models, tools, and agents with azure api management","description":"AI workloads don’t behave like traditional APIs. They combine models, tools, and agents across providers, making routing, cost control, safety, and observability harder in production. Learn how to use Azure API Management’s AI Gateway to expose endpoints, route requests, and apply policies. Through demos, go from a single request to a production-ready setup with guardrails and telemetry.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/ef7a1be3-ce2b-4442-9a91-3bc3e7eeb991","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/ef7a1be3-ce2b-4442-9a91-3bc3e7eeb991","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/ef7a1be3-ce2b-4442-9a91-3bc3e7eeb991","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/ef7a1be3-ce2b-4442-9a91-3bc3e7eeb991","speakerIds":["1768245294609build26-1757469387304001ZlCO","1768245294609build26-1778257509906001FotM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.383Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/ef7a1be3-ce2b-4442-9a91-3bc3e7eeb991","ogTitle":"Govern AI models, tools, and agents with Azure API Management","ogDescription":"AI workloads don’t behave like traditional APIs. They combine models, tools, and agents across providers, making routing, cost control, safety, and observability harder in production. Learn how to use Azure API Management’s AI Gateway to expose endpoints, route requests, and apply policies. Through demos, go from a single request to a production-ready setup with guardrails and telemetry.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778257509906001FotM","RegistrantKey":"1768245294609build26-1778257509906001FotM"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1757469387304001ZlCO","RegistrantKey":"1768245294609build26-1757469387304001ZlCO"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Governance","logicalValue":"Governance"},{"displayValue":"Tracing","logicalValue":"Tracing"},{"displayValue":"Guardrails","logicalValue":"Guardrails"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"fd21930d-2e08-4ac9-910e-714916777d17","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Jason Fisher, Jordan Geurten, Mariam Gewida, Jeffrey Sutherland","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775867300968001CRL3","langLocale":"en-US","sessionId":"1775867300968001CRL3","sessionInstanceId":"1775867300968001CRL3","sessionCode":"OD859","sessionCodeNormalized":"EN-US-OD859","title":"Hardening Windows: Security Foundations Developers Can Build On","sortTitle":"hardening windows: security foundations developers can build on","description":"Security assumptions are changing faster than most apps are built to handle. In this session, learn how Windows is raising the security foundation by reducing attack surface and preparing for the post‑quantum future. We’ll walk through Post‑Quantum Cryptography, moving authentication from NTLM to Kerberos, and stricter driver trust policies with live demos, so you know what’s changing, how it affects your apps, and what to do next.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/6a1ba2ff-a8c8-43cd-b0ee-67adf02eb6bb","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/6a1ba2ff-a8c8-43cd-b0ee-67adf02eb6bb","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/6a1ba2ff-a8c8-43cd-b0ee-67adf02eb6bb","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/6a1ba2ff-a8c8-43cd-b0ee-67adf02eb6bb","speakerIds":["1768245294609build26-1775493934666001qrTz","1768245294609build26-1776099449816001Z3iP","1768245294609build26-1776099204172001ZeYS","1768245294609build26-1776099390668001KnMV"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.902Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/6a1ba2ff-a8c8-43cd-b0ee-67adf02eb6bb","ogTitle":"Hardening Windows: Security Foundations Developers Can Build On","ogDescription":"Security assumptions are changing faster than most apps are built to handle. In this session, learn how Windows is raising the security foundation by reducing attack surface and preparing for the post‑quantum future. We’ll walk through Post‑Quantum Cryptography, moving authentication from NTLM to Kerberos, and stricter driver trust policies with live demos, so you know what’s changing, how it affects your apps, and what to do next.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1775493934666001qrTz","RegistrantKey":"1768245294609build26-1775493934666001qrTz"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776099204172001ZeYS","RegistrantKey":"1768245294609build26-1776099204172001ZeYS"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776099390668001KnMV","RegistrantKey":"1768245294609build26-1776099390668001KnMV"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776099449816001Z3iP","RegistrantKey":"1768245294609build26-1776099449816001Z3iP"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows Security","logicalValue":"Windows Security"},{"displayValue":"Platform Security","logicalValue":"Platform Security"},{"displayValue":"Post Quantum Cryptography","logicalValue":"Post Quantum Cryptography"},{"displayValue":"NTLM Deprecation","logicalValue":"NTLM Deprecation"},{"displayValue":"Kerberos Authentication","logicalValue":"Kerberos Authentication"},{"displayValue":"Driver Trust Policy","logicalValue":"Driver Trust Policy"},{"displayValue":"Attack Surface Reduction","logicalValue":"Attack Surface Reduction"}]},{"@search.score":1.0,"documentId":"9ad42ebd-f00b-4630-9699-0fa341540abe","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Amar Dhamdhere, Niko Pamboukas","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799460249001htzM","langLocale":"en-US","sessionId":"1776799460249001htzM","sessionInstanceId":"1776799460249001htzM","sessionCode":"OD828","sessionCodeNormalized":"EN-US-OD828","title":"Latest Cobalt VMs and Azure Boost enhancements​","sortTitle":"latest cobalt vms and azure boost enhancements​","description":"Learn about the latest Cobalt VMs, the use cases they support, and how they compare to our previous generation Arm based VMs. Learn how these VMs are designed for cloud-native, scale-out applications using Linux, AKS, and other open-source software. Then explore recent performance and feature enhancements for Azure Boost, as well as recently announced general compute and confidential computing VMs that rely on Azure Boost. Learn how they can support your most latency sensitive applications.​","aiDescription":"<p><strong>Introduction and Overview of Azure Cobalt:</strong> The video begins with Amar Dhamdhere introducing himself as a Project Manager from the Azure Compute VM Sizes Team and sets the context by describing Azure’s commitment to optimized virtual machine (VM) performance (<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>). He explains the significance of the Cobalt line of CPUs—Microsoft’s first in-house silicon designed specifically for the cloud. Dhamdhere highlights how Cobalt 100 revolutionized performance per cost metrics and its influence in reshaping cloud workload optimization. The agenda includes understanding customer adoption and performance gains of Cobalt 100, an introduction to Cobalt 200, performance benchmarks, VM lineup details, and deployment scenarios (<a class=\"medius-aisummary\" data-starttime=\"00:00:53\">00:00:53</a>).</p>\n\n<p><strong>Performance Success of Cobalt 100 and Customer Adoption:</strong> Amar reflects on the 2024 general availability of Azure Cobalt 100, emphasizing it as Microsoft’s first Arm-powered CPU custom-built for the cloud (<a class=\"medius-aisummary\" data-starttime=\"00:01:42\">00:01:42</a>). He shares metrics that demonstrate up to 2x .NET performance improvements, up to 20% cost reductions, and up to 45% better real-world workload performance compared to previous generations (<a class=\"medius-aisummary\" data-starttime=\"00:02:01\">00:02:01</a>). Cobalt 100’s efficiency led to reduced core demand and lower carbon footprints. Major internal services like Teams, Cosmos DB, Azure SQL, and Microsoft Defender have standardized on it, underscoring its reliability. Customer adoption significance is discussed across industries—developers using Windows on Arm, containerization pipelines, and AI agent workloads—all benefiting from high performance and agility (<a class=\"medius-aisummary\" data-starttime=\"00:03:47\">00:03:47</a>).</p>\n\n<p><strong>Launch and Capabilities of Cobalt 200:</strong> Building on Cobalt 100’s success, Amar introduces Cobalt 200 as Azure’s new leading performance CPU with up to 50% performance uplift (<a class=\"medius-aisummary\" data-starttime=\"00:05:08\">00:05:08</a>). The architecture refines design for efficiency and scales better for modern workloads such as microservices, distributed data systems, and AI engine orchestration. Benchmarks reveal consistent gains across industry-standard, Microsoft-internal, and real product workloads—all confirming per-core performance improvements. Cobalt 200’s VM families cover general, memory-optimized, and specialized workloads leveraging Azure Boost for energy efficiency (<a class=\"medius-aisummary\" data-starttime=\"00:06:41\">00:06:41</a>). The presentation emphasizes its alignment with agentic AI, with faster agent sandbox initialization, higher concurrency, and lower latency translating directly into value for developers and customers (<a class=\"medius-aisummary\" data-starttime=\"00:08:17\">00:08:17</a>).</p>\n\n<p><strong>Expanded VM Families, Tools, and Deployment Flexibility:</strong> Amar details the new Cobalt 200 VM portfolios supporting developers with DPS, EPS, and new high-memory MP and storage-focused LP series (<a class=\"medius-aisummary\" data-starttime=\"00:11:07\">00:11:07</a>). Each model is optimized for specific use cases, from microservices and web applications to large-scale analytics and caching. Cobalt 200 runs up to 70 GB/s storage throughput and supports up to 7 TB of local NVMe storage. It accommodates Linux and Windows variants and maintains full Azure feature parity, ensuring seamless migration from x86 instances. The new region rollout begins with the US and Sweden, with global availability expanding thereafter (<a class=\"medius-aisummary\" data-starttime=\"00:14:13\">00:14:13</a>). Developers are guided toward documentation, technical blogs, and migration resources to begin adopting Cobalt 200 VMs efficiently (<a class=\"medius-aisummary\" data-starttime=\"00:15:02\">00:15:02</a>).</p>\n\n<p><strong>Azure Boost Technology and RDMA Demonstration:</strong> Niko Pamboukas from the Azure Boost Team introduces the next-generation Azure Boost platform (<a class=\"medius-aisummary\" data-starttime=\"00:15:20\">00:15:20</a>). Azure Boost offloads traditional hypervisor tasks to specialized hardware to enhance IOPS, reduce maintenance overhead, and achieve up to 400 Gbps network throughput and 1M IOPS storage performance (<a class=\"medius-aisummary\" data-starttime=\"00:17:04\">00:17:04</a>). Pamboukas focuses on Guest RDMA, which allows direct memory-to-memory data transfers, bypassing typical CPU and OS layers. A demo comparing TCP vs RDMA during distributed AI inference using large LLM workloads shows up to 2.2x throughput, 3x faster time-to-first-token, and 25% token efficiency improvements (<a class=\"medius-aisummary\" data-starttime=\"00:23:18\">00:23:18</a>). He concludes with an invitation to the 2026 preview program for RDMA-enabled Azure Boost.</p>\n\n<p><strong>Intel, AMD VM Announcements and Azure Compute Security Innovations:</strong> The final segments feature Amar detailing new Intel and AMD general compute VMs powered by Azure Boost and next-gen CPUs (<a class=\"medius-aisummary\" data-starttime=\"00:24:35\">00:24:35</a>). Intel DSV7 and ESV7 series offer up to 15% performance gains with 400 Gbps networking and 20 GB/s remote storage throughput. AMD’s V7 family achieves up to 35% CPU performance and 80–130% gains on specific workloads using the latest EPYC Turin processors (<a class=\"medius-aisummary\" data-starttime=\"00:28:35\">00:28:35</a>). Vikas Bhatia then covers Azure’s compute security stack—trusted launch, confidential VMs using AMD SEV-SNP and Intel TDX, live migration for secure workloads, and Azure Integrated HSM (<a class=\"medius-aisummary\" data-starttime=\"00:30:31\">00:30:31</a>). These hardware-enforced protections ensure encryption of data in use, uphold isolation between tenants and the cloud provider, and deliver verifiable, compliant, end-to-end protection. The session concludes with resources for developers to deepen expertise in confidential computing and secure workload deployment on Azure (<a class=\"medius-aisummary\" data-starttime=\"00:39:02\">00:39:02</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/928b0d8e-af20-4b87-a5b6-8becc44e2911","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/928b0d8e-af20-4b87-a5b6-8becc44e2911","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/928b0d8e-af20-4b87-a5b6-8becc44e2911","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/928b0d8e-af20-4b87-a5b6-8becc44e2911","speakerIds":["1768245294609build26-1768580784927001elb0","1768245294609build26-1769465084231001m1qH"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/928b0d8e-af20-4b87-a5b6-8becc44e2911","ogTitle":"Latest Cobalt VMs and Azure Boost enhancements​","ogDescription":"Learn about the latest Cobalt VMs, the use cases they support, and how they compare to our previous generation Arm based VMs. Learn how these VMs are designed for cloud-native, scale-out applications using Linux, AKS, and other open-source software. Then explore recent performance and feature enhancements for Azure Boost, as well as recently announced general compute and confidential computing VMs that rely on Azure Boost. Learn how they can support your most latency sensitive applications.​","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1768580784927001elb0","RegistrantKey":"1768245294609build26-1768580784927001elb0"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1769465084231001m1qH","RegistrantKey":"1768245294609build26-1769465084231001m1qH"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"3db4fe47-d905-4293-82bb-6f1db7a17599","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776274980780001dYTL","langLocale":"en-US","sessionId":"1776274980780001dYTL","sessionInstanceId":"1776274980780001dYTL","sessionCode":"ODSP930","sessionCodeNormalized":"EN-US-ODSP930","title":"Low-latency AI for developer workflows","sortTitle":"low-latency ai for developer workflows","description":"Developer experience leaders from Cerebras and OpenAI show how low-latency inference changes coding agents, voice-driven workflows, slide generation, and workplace automation. See what you can build when responses land in seconds.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/b7bed6bb-96e2-4d0c-a4fe-463201985e85","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/b7bed6bb-96e2-4d0c-a4fe-463201985e85","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/b7bed6bb-96e2-4d0c-a4fe-463201985e85","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/b7bed6bb-96e2-4d0c-a4fe-463201985e85","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.902Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Cerebras"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/b7bed6bb-96e2-4d0c-a4fe-463201985e85","ogTitle":"Low-latency AI for developer workflows","ogDescription":"Developer experience leaders from Cerebras and OpenAI show how low-latency inference changes coding agents, voice-driven workflows, slide generation, and workplace automation. See what you can build when responses land in seconds.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"5621e33a-5c47-49fb-abfa-146a3d5fdaf6","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Suyash Kshirsagar, Kent Weare","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1779487247136001Sgb3","langLocale":"en-US","sessionId":"1779487247136001Sgb3","sessionInstanceId":"1779487247136001Sgb3","sessionCode":"OD849","sessionCodeNormalized":"EN-US-OD849","title":"Microsoft Dataverse plugin: unleashing coding agents on the enterprise","sortTitle":"microsoft dataverse plugin: unleashing coding agents on the enterprise","description":"Coding agents are powerful, but without domain tooling they hallucinate and produce broken solutions. The Dataverse plugin solves this by giving AI agents guardrailed access to tables, columns, relationships, views, security and solutions. See how a natural language request triggers multi-step provisioning, data imports and validation. All executed autonomously. We demo the plugin architecture, MCP server integration and patterns that make agent-driven Dataverse development reliable at scale.","aiDescription":"<p><strong>Introduction and Purpose:</strong> The session opens with hosts Kent Weare and Suyash Kshirsagar welcoming viewers to \"Talk to Your Terminal Using Dataverse Plugin for Coding Agent Sessions\" (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:16\">00:00:16</a>). Kent explains that while business data lives inside Microsoft Dataverse, obtaining value from it still often requires navigating portals and issuing formal queries. The new Dataverse Skills plugin aims to simplify this by letting developers, analysts, and administrators describe their intent in natural language and have it automatically executed as governed Dataverse operations (<a class=\"medius-aisummary\" data-starttime=\"00:00:21\">00:00:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>). The demonstration will show how three distinct roles — a builder, a DevOps analyst, and a systems admin — can use this single plugin to accelerate their workflows.</p>\n\n<p><strong>Builder Workflow – Maya’s Scenario:</strong> In the first demo, Maya, a business user new to Dataverse, wants to connect to her environment and create a data model for a coffee roasting business (<a class=\"medius-aisummary\" data-starttime=\"00:01:12\">00:01:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>). Without knowing her organization URL, she installs the Dataverse Skills plugin, connects using Microsoft sign-in, and has the agent automatically detect and verify her environment (<a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:06\">00:02:06</a>). Using a single natural-language prompt, Maya generates a full data schema for tracking bean batches, roasts, orders, and quality checks, complete with lookups, relationships, and a pre-built solution container (<a class=\"medius-aisummary\" data-starttime=\"00:02:13\">00:02:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>). She then uses subsequent prompts to load actual spreadsheets into her Dataverse tables, rely on business keys to resolve relationships, and validate imported data. Thanks to the Dataverse plugin’s skill orchestration, Maya completes a full working app setup, including validation queries and data consistency checks, entirely through conversational steps (<a class=\"medius-aisummary\" data-starttime=\"00:03:53\">00:03:53</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:37\">00:05:37</a>).</p>\n\n<p><strong>Analyst Workflow – Riya’s Scenario:</strong> The next segment features Riya, a DevOps analyst at Zava Coffee, showing how operational users use the same plugin to access CRM data directly from their terminals (<a class=\"medius-aisummary\" data-starttime=\"00:05:46\">00:05:46</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:12\">00:06:12</a>). After connecting to Dataverse naturally via prompt-based authentication, Riya queries CRM entities like accounts, contacts, and opportunities with simple language, replacing the need for technical FetchXML or manual portal work (<a class=\"medius-aisummary\" data-starttime=\"00:06:22\">00:06:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:58\">00:06:58</a>). Examples include finding \"Carlos’s open opportunities over $100,000 closing this quarter\" and \"cafes in Portland that haven’t reordered in 30 days\" — operations that ordinarily take many minutes are now completed instantly by the agent’s skill system (<a class=\"medius-aisummary\" data-starttime=\"00:07:03\">00:07:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:50\">00:08:50</a>). Riya also adds notes, marks activities complete, and logs calls on behalf of others with plain English prompts, while the plugin automatically determines record owners, related entities, and relationships (<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:44\">00:11:44</a>). This segment demonstrates how data retrieval, updates, and task tracking follow identical conversational patterns across users and contexts.</p>\n\n<p><strong>Administrator Workflow – Amara’s Scenario:</strong> The final demonstration shows Amara, a system administrator, using the same Dataverse plugin for complex configuration, security, and governance tasks (<a class=\"medius-aisummary\" data-starttime=\"00:12:02\">00:12:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:14\">00:12:14</a>). After confirming her privileges, Amara issues prompts to define business units for multiple regions, create roles for sales, warehouse, and leadership users, set field-level security for sensitive fields, and assign access team templates, all expressed in natural language (<a class=\"medius-aisummary\" data-starttime=\"00:13:18\">00:13:18</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:05\">00:15:05</a>). The agent automatically performs prerequisite actions such as enabling field security, binding profiles, and packaging all resources into a secure, versioned solution. Amara then validates her configuration through impersonation checks that confirm read/write capabilities across roles and units, followed by sharing records cross-region using team templates without manual ID entry (<a class=\"medius-aisummary\" data-starttime=\"00:16:14\">00:16:14</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:37\">00:17:37</a>). She finalizes by enabling and verifying auditing of sensitive fields so reads and writes are fully logged. In minutes, Amara accomplishes multi-layer Dataverse administration that previously required manual and error-prone portal steps (<a class=\"medius-aisummary\" data-starttime=\"00:18:01\">00:18:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:03\">00:19:03</a>).</p>\n\n<p><strong>Conclusion and Call to Action:</strong> Kent and Suyash close the session by reinforcing the power of a unified conversational interface across roles and disciplines (<a class=\"medius-aisummary\" data-starttime=\"00:19:08\">00:19:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:16\">00:19:16</a>). They emphasize that builders, analysts, and administrators can all work faster and more securely using the same Dataverse plugin, which turns natural language directly into managed operational outcomes. Viewers are invited to scan the on-screen QR code for more details and hands-on access (<a class=\"medius-aisummary\" data-starttime=\"00:19:19\">00:19:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:32\">00:19:32</a>), marking the end of the demonstration and highlighting Microsoft’s vision for AI-assisted Dataverse management.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/961026f5-a22a-4c0f-a9cb-80b962beb540","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/961026f5-a22a-4c0f-a9cb-80b962beb540","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/961026f5-a22a-4c0f-a9cb-80b962beb540","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/961026f5-a22a-4c0f-a9cb-80b962beb540","speakerIds":["1768245294609build26-1779487350828001URNf","1768245294609build26-1779487323974001SKUq"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:59.237Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/961026f5-a22a-4c0f-a9cb-80b962beb540","ogTitle":"Microsoft Dataverse plugin: unleashing coding agents on the enterprise","ogDescription":"Coding agents are powerful, but without domain tooling they hallucinate and produce broken solutions. The Dataverse plugin solves this by giving AI agents guardrailed access to tables, columns, relationships, views, security and solutions. See how a natural language request triggers multi-step provisioning, data imports and validation. All executed autonomously. We demo the plugin architecture, MCP server integration and patterns that make agent-driven Dataverse development reliable at scale.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"Python","logicalValue":"Python"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1779487323974001SKUq","RegistrantKey":"1768245294609build26-1779487323974001SKUq"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1779487350828001URNf","RegistrantKey":"1768245294609build26-1779487350828001URNf"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"GitHub Copilot CLI","logicalValue":"GitHub Copilot CLI"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Developer Languages","logicalValue":"Developer Languages"}]},{"@search.score":1.0,"documentId":"235b12d1-afe2-49e7-af52-d7d79dee5bb8","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Erika Heidi","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777997146800001i1Rc","langLocale":"en-US","sessionId":"1777997146800001i1Rc","sessionInstanceId":"1777997146800001i1Rc","sessionCode":"ODSP938","sessionCodeNormalized":"EN-US-ODSP938","title":"Mitigate software supply chain risks in Github Actions","sortTitle":"mitigate software supply chain risks in github actions","description":"GitHub Actions workflows are a prime target for supply chain attacks—from malicious pull requests exfiltrating secrets to compromised dependencies and hijacked tags. In this 15-minute session, we'll walk through practical, high-impact strategies to protect your CI/CD pipelines, including how to inspect repos for hidden risks, shrink your attack surface, and lock down what your workflows can actually do. Leave with a concrete checklist to harden your workflows today. Stay safe!","aiDescription":"<p><strong>Introduction and Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Erika Heidi, Staff DevRel Engineer at Chainguard, introduces herself and explains that she will present five key strategies to mitigate software supply chain risks in GitHub Actions. The talk begins by emphasizing the importance of inspecting repositories for insecure defaults and common bad practices such as using pull request targets with head code execution, broad-scope personal access tokens (PATs), direct shell execution, and actions pinned by tag (<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>). Erika suggests using GitHub Copilot to automatically evaluate workflows and detect vulnerabilities in a repository (<a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:02\">00:01:02</a>), noting it can identify numerous issues effectively.</p>\n\n<p><strong>Insecure Workflow Demonstration and Protection Measures:</strong> Erika walks through a test repository intentionally configured with insecure settings to demonstrate how secrets can be exfiltrated <a class=\"medius-aisummary\" data-starttime=\"00:01:16\">00:01:16</a>. She explains the risk of pull request targets that execute code in the context of the main branch, exposing secrets and environment variables to potential attackers (<a class=\"medius-aisummary\" data-starttime=\"00:01:38\">00:01:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>). Erika references incidents like the Trivy exploitation, where tags were rewritten to point to malicious commits (<a class=\"medius-aisummary\" data-starttime=\"00:03:07\">00:03:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:24\">00:03:24</a>). She highlights the importance of protecting the main branch and release tags to prevent unauthorized pushes and tag rewrites that could propagate malicious code. These settings, she states, need to be manually enabled in repository configuration (<a class=\"medius-aisummary\" data-starttime=\"00:04:03\">00:04:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:33\">00:04:33</a>).</p>\n\n<p><strong>Minimizing Attack Surface and Using Secure Images:</strong> The second major strategy, beginning at <a class=\"medius-aisummary\" data-starttime=\"00:04:37\">00:04:37</a>, is to minimize the attack surface. Erika advises removing unnecessary components from runtime environments to reduce potential entry points for exploitation. She explains that direct and transitive dependencies can serve as weak links in the security chain (<a class=\"medius-aisummary\" data-starttime=\"00:05:04\">00:05:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:15\">00:05:15</a>). The use of minimal container images is recommended to avoid excessive packages and vulnerabilities. Chainguard builds all packages from source, ensuring up-to-date, patched images with remarkably fewer CVEs compared to standard images from Docker Hub (<a class=\"medius-aisummary\" data-starttime=\"00:05:59\">00:05:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:26\">00:06:26</a>). This approach provides a simple yet powerful improvement in security posture.</p>\n\n<p><strong>Pulling from Trusted Sources and the Risks of Public Registries:</strong> Erika transitions to the third tip at <a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>, focusing on fetching dependencies only from trusted sources. She reports that about 98% of malware is introduced during build and distribution times, bypassing source code review (<a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:25\">00:07:25</a>). Public registries, although useful for sharing libraries, often lack sufficient security measures to prevent artifact tampering or “ghost releases,” which introduce malicious payloads during build processes (<a class=\"medius-aisummary\" data-starttime=\"00:07:50\">00:07:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:33\">00:08:33</a>). She describes how attackers exploit CI/CD pipelines and developer machines to gain credentials and spread infections laterally. Using Chainguard’s secure library repositories for Python, Java, and JavaScript ensures verified, tamper-proof builds without pre- or post-install scripts (<a class=\"medius-aisummary\" data-starttime=\"00:09:21\">00:09:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>).</p>\n\n<p><strong>Pin by Digest and Automated Updates:</strong> The fourth principle at <a class=\"medius-aisummary\" data-starttime=\"00:10:18\">00:10:18</a> discusses replacing tag-based versioning with digest-based pinning. A digest, Erika explains, is a unique cryptographic hash identifying a specific build, ensuring users execute known code every time (<a class=\"medius-aisummary\" data-starttime=\"00:10:33\">00:10:33</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:48\">00:10:48</a>). To keep digests updated safely, she introduces Digestabot—a free, open-source Chainguard tool that automatically sends pull requests with new digests when updated versions are available (<a class=\"medius-aisummary\" data-starttime=\"00:11:05\">00:11:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:24\">00:11:24</a>). This step helps maintain continuous workflow security without relying on vulnerable tags.</p>\n\n<p><strong>Managing Access Tokens and Final Recap:</strong> The last strategy, starting at <a class=\"medius-aisummary\" data-starttime=\"00:12:03\">00:12:03</a>, concerns banning long-lived PATs. Erika warns that these tokens, when leaked, grant attackers extensive control over entire organizations—as in the Trivy incident (<a class=\"medius-aisummary\" data-starttime=\"00:12:28\">00:12:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:12:36\">00:12:36</a>). She recommends short-lived tokens managed through Octo-STS, a free GitHub application implementing temporary credentials that expire quickly using Sigstore and Cosign principles (<a class=\"medius-aisummary\" data-starttime=\"00:13:08\">00:13:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:27\">00:13:27</a>). Erika concludes with a TLDR at <a class=\"medius-aisummary\" data-starttime=\"00:13:49\">00:13:49</a> summarizing the five actions—inspect repositories, minimize attack surfaces, pull from trusted sources, pin by digest, and ban long-lived tokens. She ends the presentation at <a class=\"medius-aisummary\" data-starttime=\"00:15:06\">00:15:06</a> by encouraging viewers to stay secure and vigilant.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/374e6618-f291-4b4a-8e63-fe07d8bf854a","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/374e6618-f291-4b4a-8e63-fe07d8bf854a","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/374e6618-f291-4b4a-8e63-fe07d8bf854a","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/374e6618-f291-4b4a-8e63-fe07d8bf854a","speakerIds":["1768245294609build26-1780348431411001xOTh"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.767Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Chainguard"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/374e6618-f291-4b4a-8e63-fe07d8bf854a","ogTitle":"Mitigate software supply chain risks in Github Actions","ogDescription":"GitHub Actions workflows are a prime target for supply chain attacks—from malicious pull requests exfiltrating secrets to compromised dependencies and hijacked tags. In this 15-minute session, we'll walk through practical, high-impact strategies to protect your CI/CD pipelines, including how to inspect repos for hidden risks, shrink your attack surface, and lock down what your workflows can actually do. Leave with a concrete checklist to harden your workflows today. Stay safe!","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"},{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1780348431411001xOTh","RegistrantKey":"1768245294609build26-1780348431411001xOTh"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"1fab283e-3c37-42a3-8d6b-3729334f016e","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gaurav Seth, Andrew Westgarth","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778013174628001szVP","langLocale":"en-US","sessionId":"1778013174628001szVP","sessionInstanceId":"1778013174628001szVP","sessionCode":"OD801","sessionCodeNormalized":"EN-US-OD801","title":"Modernize intelligent apps and agents with .NET that scale as you grow","sortTitle":"modernize intelligent apps and agents with .net that scale as you grow","description":".NET helps you build high performance services, add AI where it matters, and stay productive across the stack. In this session, we show how using Managed Instance on Azure App Service, you can modernize legacy .NET apps without code rewrites, remove migration blockers, retain dependencies, and move to a scalable PaaS foundation while reducing cost and accelerating innovation.","aiDescription":"<p><strong>Introduction and Session Overview:</strong> The session begins with Andrew Westgarth welcoming viewers to the presentation on modernizing intelligent apps and agents with .NET that scale on Azure (<a class=\"medius-aisummary\" data-starttime=\"00:00:07\">00:00:07</a>). Joined by co-presenter Gaurav Seth, Andrew outlines that both are product managers on Azure App Service and will explore recent innovations that accelerate cloud migration and AI integration. The agenda covers app modernization challenges, the previously launched <em>Managed Instance on Azure App Service</em> from Ignite (<a class=\"medius-aisummary\" data-starttime=\"00:00:58\">00:00:58</a>), the newly introduced <em>Built-In MCP</em> at Build (<a class=\"medius-aisummary\" data-starttime=\"00:01:18\">00:01:18</a>), the GitHub Copilot modernization tooling, and a live demonstration by Gaurav followed by closing resources and insights on cloud transformation.</p>\n\n<p><strong>Modernization Challenges and Strategic Needs:</strong> Andrew discusses critical barriers organizations face when transitioning applications to cloud environments (<a class=\"medius-aisummary\" data-starttime=\"00:02:14\">00:02:14</a>). He outlines dependencies on legacy systems, third-party libraries, and Windows OS components such as MSMQ and GDI libraries that complicate modernization. Other major issues include handling stateful versus stateless architecture for scalability (<a class=\"medius-aisummary\" data-starttime=\"00:03:31\">00:03:31</a>), managing configuration and secrets securely (<a class=\"medius-aisummary\" data-starttime=\"00:04:12\">00:04:12</a>), and dealing with lengthy migration timelines that impact ROI (<a class=\"medius-aisummary\" data-starttime=\"00:04:39\">00:04:39</a>). The discussion sets the stage for how Managed Instance and integrated AI solutions simplify these areas—offering centralized governance, reduced operational overhead, and faster deployment without excessive code rewrites.</p>\n\n<p><strong>Managed Instance on Azure App Service:</strong> Andrew introduces the <em>Managed Instance on Azure App Service</em> preview as a foundational element for easing migration (<a class=\"medius-aisummary\" data-starttime=\"00:05:06\">00:05:06</a>). This service combines the flexibility of virtual machines with the scalability of platform-as-a-service, addressing Windows and third-party dependency challenges while supporting both horizontal and vertical scaling. It offers built-in security via Managed Identity and integration with services like Key Vault and SQL (<a class=\"medius-aisummary\" data-starttime=\"00:05:59\">00:05:59</a>). With minimal or zero code changes, developers can migrate apps while maintaining access to mapped drives, registry keys, and configuration compatibility (<a class=\"medius-aisummary\" data-starttime=\"00:07:08\">00:07:08</a>). Managed Instance also enables direct RDP access through Bastion for troubleshooting, dynamic scalability, automatic maintenance, and incremental modernization for continuous improvement after migration (<a class=\"medius-aisummary\" data-starttime=\"00:10:04\">00:10:04</a>).</p>\n\n<p><strong>Introducing Built-In MCP and Agentic Patterns:</strong> The session reveals a major new capability—<em>Built-In MCP</em>—unveiled at Build (<a class=\"medius-aisummary\" data-starttime=\"00:13:03\">00:13:03</a>). This feature allows APIs to function as AI tools accessible to intelligent agents using secure, language-agnostic endpoints that support .NET, Java, Node.js, and Python. The presenters describe how existing applications can expose APIs seamlessly for AI consumption through MCP interfaces (<a class=\"medius-aisummary\" data-starttime=\"00:13:12\">00:13:12</a>). Andrew also introduces agentic application patterns (<a class=\"medius-aisummary\" data-starttime=\"00:15:02\">00:15:02</a>)—modular approaches where specialized agents (e.g., sales or shipping agents) coordinate tasks orchestrated by AI—allowing dynamic scaling, hybrid integration with legacy systems, and intelligence-driven automation across enterprise workloads.</p>\n\n<p><strong>GitHub Copilot Modernization Tooling and Live Demo:</strong> Gaurav Seth demonstrates how GitHub Copilot aids modernization from within Visual Studio (<a class=\"medius-aisummary\" data-starttime=\"00:19:29\">00:19:29</a>). He runs an AI-powered assessment of a legacy ASP.NET web app, which identifies migration readiness and dependency issues like registry or local storage handling (<a class=\"medius-aisummary\" data-starttime=\"00:21:14\">00:21:14</a>). Using Copilot’s built-in recommendations, developers can target Azure App Service Managed Instance to overcome sandbox limitations and accelerate deployment. Gaurav walks through publishing an app, configuring installation scripts for custom components, managing storage mounts and registries secured via Key Vault, and connecting through Bastion RDP for live system access (<a class=\"medius-aisummary\" data-starttime=\"00:36:36\">00:36:36</a>). He then showcases turning an existing API into an MCP server that integrates directly with GitHub Copilot and shows how observability data for agents is collected and visualized through App Insights (<a class=\"medius-aisummary\" data-starttime=\"00:45:01\">00:45:01</a>).</p>\n\n<p><strong>Conclusion and Key Resources:</strong> In closing, Andrew summarizes the session’s key lessons and directs attendees to explore documentation and blogs for <em>App Service @ Build</em> 2026 (<a class=\"medius-aisummary\" data-starttime=\"00:46:39\">00:46:39</a>). He highlights learning paths on agentic app development, Built-In MCP for AI integration, and the Managed Instance overview. Additional materials cover GitHub Copilot modernization strategies and examples of automated IIS migrations leveraging agent-based workflows. Andrew concludes by reiterating that Azure App Service now enables enterprises to move, modernize, and intelligently extend their .NET applications with unparalleled agility, security, and AI-connected innovation (<a class=\"medius-aisummary\" data-starttime=\"00:49:00\">00:49:00</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/d14b3177-b1c2-481c-9094-4fe12dc84b75","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/d14b3177-b1c2-481c-9094-4fe12dc84b75","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/d14b3177-b1c2-481c-9094-4fe12dc84b75","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/d14b3177-b1c2-481c-9094-4fe12dc84b75","speakerIds":["1768245294609build26-1765534448626001C3HQ","1768245294609build26-1778642279845001saaB"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.769Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/d14b3177-b1c2-481c-9094-4fe12dc84b75","ogTitle":"Modernize intelligent apps and agents with .NET that scale as you grow","ogDescription":".NET helps you build high performance services, add AI where it matters, and stay productive across the stack. In this session, we show how using Managed Instance on Azure App Service, you can modernize legacy .NET apps without code rewrites, remove migration blockers, retain dependencies, and move to a scalable PaaS foundation while reducing cost and accelerating innovation.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778642279845001saaB","RegistrantKey":"1768245294609build26-1778642279845001saaB"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1765534448626001C3HQ","RegistrantKey":"1768245294609build26-1765534448626001C3HQ"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"f5157047-6c30-4157-8776-6bb65c367212","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ayan Gupta","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778258199294001WdQJ","langLocale":"en-US","sessionId":"1778258199294001WdQJ","sessionInstanceId":"1778258199294001WdQJ","sessionCode":"OD871","sessionCodeNormalized":"EN-US-OD871","title":"Modernize Java apps with AI​","sortTitle":"modernize java apps with ai​","description":"Modernizing legacy Java apps can feel overwhelming—unfamiliar code, outdated frameworks, and fear of breaking what works. This series shows how to use GitHub Copilot to turn modernization into a structured, AI-assisted workflow. Follow Ayan Gupta, Sandra Ahlgrimm, and Yoshio Terada through each phase—from assessment to cloud deployment on Azure—featuring a real migration from Java 5/Struts to Java 21/Spring Boot.​","aiDescription":"<p><strong>Series Introduction and Background:</strong> At the start of the series (<a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:37\">01:37</a>), host Ian introduces \"Modernize Java Apps with AI\" from Microsoft HQ in Redmond. He begins with a cautionary story referencing the Louvre heist, explaining how aging systems and delayed modernization expose organizations to risk. This sets the theme: modernization isn’t just upgrading technology—it’s ensuring resilience. Ian frames modernization as a proactive process, using the GitHub Copilot Modernize extension that identifies issues before they escalate. The goal of this new video series is presented as exploring how AI-driven Copilot enables enterprises to modernize Java applications efficiently through steps like assessment, upgrading frameworks, migration, containerization, and automation. The introduction places modernization as a critical enabler for scaling AI adoption in enterprise environments.</p>\n\n<p><strong>Modernization Framework and Core Workflow:</strong> Continuing from (<a class=\"medius-aisummary\" data-starttime=\"01:38\">01:38</a>–<a class=\"medius-aisummary\" data-starttime=\"07:17\">07:17</a>), Ian explains the structured modernization loop—assessment, upgrade, migration, testing, and deployment—with AI supercharging each stage. The video details how GitHub Copilot integrates with developer workflows via portfolio assessment tools like Azure Migrate and Doctor Migrate to automatically create actionable GitHub issues that developers can pick up inside IDEs such as VS Code, Intellij, or Visual Studio. Automation reduces repetitive modernization tasks—like dependency upgrades, containerization, fixing vulnerabilities (CVEs), and configuration updates—from days to hours. This section emphasizes modernization as iterative and continuous, highlighting AI agents executing in defined loops, ensuring issues don’t recur, and maintaining developer oversight throughout the process.</p>\n\n<p><strong>Assessment and Planning Phase with Copilot:</strong> The next episode (<a class=\"medius-aisummary\" data-starttime=\"08:00\">08:00</a>–<a class=\"medius-aisummary\" data-starttime=\"16:44\">16:44</a>) demonstrates the application assessment stage—the first step toward modernization. Ian uses a sample enterprise e-commerce app (a ski shop) built on outdated Java 1.5 and Struts frameworks to illustrate tech debt scenarios. With GitHub Copilot’s assessment extension, he shows how AI identifies frameworks, configurations, and upgrade paths through static analysis and rule inspection, displaying results live in the “Mission Control” UI. The episode covers how issues are categorized into cloud readiness, runtime compatibility, and security risks, creating transparent reports and allowing developers to build custom, repeatable workflows. By turning complex unknown codebases into clear, ordered tasks, modernization becomes developer-driven rather than daunting.</p>\n\n<p><strong>Executing Upgrades and Handling Dependencies with AI Agents:</strong> In the following episode hosted by Sandra (<a class=\"medius-aisummary\" data-starttime=\"16:45\">16:45</a>–<a class=\"medius-aisummary\" data-starttime=\"41:21\">41:21</a>), viewers see a live example of upgrading a Spring Boot application from Java 8 to Java 25 using Copilot modernization. Sandra walks through using Intellij with AI agents to automatically generate and execute upgrade plans in milestones—checking runtime environments, applying open-source rewrite tools, scanning CVEs, running tests, and validating behaviors after upgrades. The demo shows active debugging, resolution of version compatibility issues, and Copilot’s ability to fix failing unit tests autonomously while maintaining human oversight. Eventually, the upgraded app runs successfully on port 8080 with all tests passing. This episode demonstrates AI-assisted modernization at work—transforming months of manual upgrades into minutes while ensuring quality and security compliance.</p>\n\n<p><strong>Team Experiences and Customization Workflow:</strong> The series also showcases contributions from Yoshio (<a class=\"medius-aisummary\" data-starttime=\"41:27\">41:27</a>–<a class=\"medius-aisummary\" data-starttime=\"57:50\">57:50</a>) and others, who explain modernization principles like the “six R” approaches and introduce task customization. Using GitHub Copilot’s modernization for Java, Yoshio migrates a legacy Struts monolith to Spring in two days instead of weeks, underscoring AI’s ability to reduce workload. Later demonstrations show how developers can define constraints and create explicit modernization tasks—what to modify, what to exclude, and what requires manual verification. Copilot’s workflow ensures every modernization action remains transparent and developer-controlled, identifying CVEs or test failures like Mockito incompatibility without changing files silently. This segment reinforces the principle of safe modernization under strict version control and human review.</p>\n\n<p><strong>Containerization, Deployment, and CLI Automation:</strong> In further episodes (<a class=\"medius-aisummary\" data-starttime=\"58:01\">58:01</a>–<a class=\"medius-aisummary\" data-starttime=\"78:40\">78:40</a>), Ian and team demonstrate containerizing the upgraded Java app and deploying it to Azure. GitHub Copilot generates Docker assets, validates environments, and builds deployment plans using Azure Bicep templates. The application is tested locally with H2 database before deployment, and once deployed via container apps, it runs live on Azure. Subsequently, modernization moves to command-line environments, where multiple repositories are assessed and upgraded in parallel using Copilot CLI agents. Outputs like plans, diffs, and reports are pushed to GitHub for review—proving modernization can scale across dozens of repos seamlessly. This segment expands modernization from singular workflows in IDEs to automated, iterative cloud-scale processes embedded into CI/CD pipelines.</p>\n\n<p><strong>Series Conclusion and Core Takeaways:</strong> The concluding episode (<a class=\"medius-aisummary\" data-starttime=\"78:56\">78:56</a>–<a class=\"medius-aisummary\" data-starttime=\"81:47\">81:47</a>) summarizes the journey—covering assessment, planning, execution, upgrades, migration, containerization, deployment, and scaling through AI assistance. It stresses that modernization and migration are not one-off projects but continuous, reviewable engineering practices integrated directly into developer workflows. The Copilot modernization loop provides repeatability, efficiency, transparency, and control, allowing developers to focus on architecture and innovation while AI handles mechanical tasks. The series closes by encouraging viewers to integrate modernization pipelines into their existing tools and CI/CD environments for sustainable maintenance and AI readiness, thus empowering organizations to evolve securely into the future.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/202a3fe5-c390-49f0-bd58-ca6cde61a6fe","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/202a3fe5-c390-49f0-bd58-ca6cde61a6fe","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/202a3fe5-c390-49f0-bd58-ca6cde61a6fe","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/202a3fe5-c390-49f0-bd58-ca6cde61a6fe","speakerIds":["1768245294609build26-1767923733240001J5fY"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.774Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/202a3fe5-c390-49f0-bd58-ca6cde61a6fe","ogTitle":"Modernize Java apps with AI​","ogDescription":"Modernizing legacy Java apps can feel overwhelming—unfamiliar code, outdated frameworks, and fear of breaking what works. This series shows how to use GitHub Copilot to turn modernization into a structured, AI-assisted workflow. Follow Ayan Gupta, Sandra Ahlgrimm, and Yoshio Terada through each phase—from assessment to cloud deployment on Azure—featuring a real migration from Java 5/Struts to Java 21/Spring Boot.​","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[{"displayValue":"Java","logicalValue":"Java"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1767923733240001J5fY","RegistrantKey":"1768245294609build26-1767923733240001J5fY"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"GitHub","logicalValue":"GitHub"},{"displayValue":"App Mod","logicalValue":"App Mod"}]},{"@search.score":1.0,"documentId":"266402f5-d9ab-417d-93a8-1493f91eb629","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396210720001zt5r","langLocale":"en-US","sessionId":"1774396210720001zt5r","sessionInstanceId":"1774396210720001zt5r","sessionCode":"ODSP907","sessionCodeNormalized":"EN-US-ODSP907","title":"Monitor GenAI applications beyond golden signals","sortTitle":"monitor genai applications beyond golden signals","description":"The golden signals of monitoring (Latency, Errors, Traffic, Saturation) remain foundational, but for GenAI applications they leave critical blind spots. A 200 OK with low latency doesn't tell you the response hallucinated, leaked PII, or cost more than what it should.","aiDescription":"<p><strong>Introduction to GenAI Monitoring:</strong> The video begins with a welcome to \"Beyond Golden Signals: Monitoring in the Age of Generative AI\" <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>. The presenter explains how generative AI (GenAI) systems differ from traditional software, identifying four fundamental shifts: non-deterministic behavior, variable cost structures, new attack vectors, and subjective quality <a class=\"medius-aisummary\" data-starttime=\"00:00:15\">00:00:15</a>. These differences make legacy monitoring strategies insufficient, prompting the need for new observability frameworks tailored to AI-driven systems.</p>\n\n<p><strong>Revisiting the Golden Signals (LETS):</strong> The speaker introduces the traditional \"Golden Signals\"—Latency, Errors, Traffic, and Saturation (LETS)—as the foundation of monitoring <a class=\"medius-aisummary\" data-starttime=\"00:00:37\">00:00:37</a>. For latency, traditional metrics such as median, P95, and P99 are useful, but GenAI requires more granular tracking across multiple pipeline stages like RAG retrieval time, LLM response time, and total request duration <a class=\"medius-aisummary\" data-starttime=\"00:00:45\">00:00:45</a>. Error tracking must go beyond HTTP status codes to include model errors such as context length limits and safety filter triggers <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>. Traffic monitoring shifts to segmenting by model, feature, and user type to manage dynamic demand and rate limits <a class=\"medius-aisummary\" data-starttime=\"00:02:41\">00:02:41</a>. Finally, saturation now emphasizes GPU utilization and API rate limits as primary bottlenecks in AI systems <a class=\"medius-aisummary\" data-starttime=\"00:04:05\">00:04:05</a>.</p>\n\n<p><strong>New Dimension 1 - Cost Monitoring:</strong> As the first new dimension of AI observability, cost monitoring is highlighted as a necessity <a class=\"medius-aisummary\" data-starttime=\"00:04:34\">00:04:34</a>. GenAI brings unpredictable cost dynamics due to model use and token-based billing. The video identifies three primary sources of cost escalation: token creep, where larger context windows inflate costs; model drift, where switching to more expensive models increases expenditure; and uncached calls, which repeat identical costly prompts unnecessarily <a class=\"medius-aisummary\" data-starttime=\"00:05:10\">00:05:10</a>. A strong cost attribution system based on tagging—by feature, user, model, and endpoint—is recommended to identify cost drivers, enable chargeback accuracy, detect abuse, and maintain budget control <a class=\"medius-aisummary\" data-starttime=\"00:06:26\">00:06:26</a>.</p>\n\n<p><strong>New Dimension 2 - Safety and Security:</strong> The discussion transitions to safety and security monitoring <a class=\"medius-aisummary\" data-starttime=\"00:07:37\">00:07:37</a>. The GenAI threat landscape includes critical risks like PII leakage and data exfiltration, high-risk attacks such as prompt injection and jailbreaking, and medium-risk issues like denial-of-wallet and model extraction <a class=\"medius-aisummary\" data-starttime=\"00:07:53\">00:07:53</a>. Effective monitoring requires metrics that detect non-traditional security breaches—tracking prompt injection rates, PII detection rates, moderation scores, and jailbreak attempts <a class=\"medius-aisummary\" data-starttime=\"00:08:26\">00:08:26</a>. These indicators help identify misuse or malicious manipulation of large language models before they cause widespread damage.</p>\n\n<p><strong>New Dimension 3 - Quality Monitoring:</strong> The final new dimension—quality monitoring—addresses the most subjective yet critical aspect of GenAI <a class=\"medius-aisummary\" data-starttime=\"00:08:59\">00:08:59</a>. Traditional uptime metrics cannot measure whether AI outputs are helpful, accurate, or complete. Quality assessment is difficult due to lack of ground truth, contextual dependency, hallucinations, and user satisfaction variability <a class=\"medius-aisummary\" data-starttime=\"00:09:17\">00:09:17</a>. Six quality metrics are proposed: hallucination rate, relevance score, user satisfaction, answer completeness, retrieval quality, and response coherence <a class=\"medius-aisummary\" data-starttime=\"00:10:11\">00:10:11</a>. These enable a multidimensional understanding of performance beyond binary success indicators.</p>\n\n<p><strong>Conclusion and Platform Overview:</strong> In closing, the speaker unifies the traditional LETS framework with the three new dimensions—Cost, Safety, and Quality—to form a comprehensive GenAI observability model <a class=\"medius-aisummary\" data-starttime=\"00:10:45\">00:10:45</a>. This integration enables visibility into system health, financial control, security posture, and model performance <a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>. Datadog’s LLM Observability platform is introduced, offering monitoring for latency, errors, token usage, costs, quality metrics, and security compliance in one interface <a class=\"medius-aisummary\" data-starttime=\"00:11:28\">00:11:28</a>. The session concludes by reaffirming that traditional metrics still matter, but alone they offer an incomplete picture—true GenAI monitoring must address hallucinations, data leakage, and uncontrolled spending simultaneously <a class=\"medius-aisummary\" data-starttime=\"00:12:27\">00:12:27</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/7326ca8e-1c9d-4823-a760-6bb36d6212e0","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/7326ca8e-1c9d-4823-a760-6bb36d6212e0","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/7326ca8e-1c9d-4823-a760-6bb36d6212e0","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/7326ca8e-1c9d-4823-a760-6bb36d6212e0","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.639Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Datadog"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/7326ca8e-1c9d-4823-a760-6bb36d6212e0","ogTitle":"Monitor GenAI applications beyond golden signals","ogDescription":"The golden signals of monitoring (Latency, Errors, Traffic, Saturation) remain foundational, but for GenAI applications they leave critical blind spots. A 200 OK with low latency doesn't tell you the response hallucinated, leaked PII, or cost more than what it should.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Azure Monitor","logicalValue":"Azure Monitor"},{"displayValue":"Cost Management","logicalValue":"Cost Management"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Semantic Kernel","logicalValue":"Semantic Kernel"},{"displayValue":"Vector Embeddings","logicalValue":"Vector Embeddings"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"AI Toolkit","logicalValue":"AI Toolkit"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Containment","logicalValue":"Containment"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Skills","logicalValue":"Skills"},{"displayValue":"Claws","logicalValue":"Claws"},{"displayValue":"Openclaw","logicalValue":"Openclaw"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"0429e5d1-4ec5-4df4-a6bb-309831ed85fa","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Ram Kakani, Jeff Smith","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774543915443001njLx","langLocale":"en-US","sessionId":"1774543915443001njLx","sessionInstanceId":"1774543915443001njLx","sessionCode":"ODSP919","sessionCodeNormalized":"EN-US-ODSP919","title":"Move from data to intelligence with Oracle MCP and Microsoft IQ","sortTitle":"move from data to intelligence with oracle mcp and microsoft iq","description":"Make the leap from disconnected data to intelligent, AI-driven workflows by combining Oracle managed MCP Servers with Microsoft IQ. See how MCP enables connectivity to Oracle Database@Azure, while Microsoft IQ—Work IQ, Fabric IQ, and Foundry IQ—bring context, reasoning, and orchestration to enterprise data. Together, they power agentic experiences with Oracle handling infrastructure and operations, accelerating time to value, and delivering enterprise-grade solutions at scale.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/8e64f47d-7993-47ae-a196-8100b99f5183","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/8e64f47d-7993-47ae-a196-8100b99f5183","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/8e64f47d-7993-47ae-a196-8100b99f5183","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/8e64f47d-7993-47ae-a196-8100b99f5183","speakerIds":["1768245294609build26-1768000325149001W0Ua","1768245294609build26-1778294630241001cb9K"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.842Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Oracle"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/8e64f47d-7993-47ae-a196-8100b99f5183","ogTitle":"Move from data to intelligence with Oracle MCP and Microsoft IQ","ogDescription":"Make the leap from disconnected data to intelligent, AI-driven workflows by combining Oracle managed MCP Servers with Microsoft IQ. See how MCP enables connectivity to Oracle Database@Azure, while Microsoft IQ—Work IQ, Fabric IQ, and Foundry IQ—bring context, reasoning, and orchestration to enterprise data. Together, they power agentic experiences with Oracle handling infrastructure and operations, accelerating time to value, and delivering enterprise-grade solutions at scale.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778294630241001cb9K","RegistrantKey":"1768245294609build26-1778294630241001cb9K"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1768000325149001W0Ua","RegistrantKey":"1768245294609build26-1768000325149001W0Ua"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"API","logicalValue":"API"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"App Integration","logicalValue":"App Integration"}]},{"@search.score":1.0,"documentId":"e2db0d78-1fb8-40b7-804f-abe79729c825","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"David Haikney","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774394693866001oZgJ","langLocale":"en-US","sessionId":"1774394693866001oZgJ","sessionInstanceId":"1774394693866001oZgJ","sessionCode":"ODSP900","sessionCodeNormalized":"EN-US-ODSP900","title":"Performance tuning on Cobalt with Arm Performix​","sortTitle":"performance tuning on cobalt with arm performix​","description":"Learn how to identify code hotspots, interpret performance data, and move from raw metrics to concrete improvements. The session also explores advanced command-line usage and shows how Arm Performix can be integrated into automated and agentic workflows for continuous performance analysis.​ By the end of the session, you’ll be able to apply Performix in real-world development pipelines to continuously monitor, diagnose, and optimize application performance.","aiDescription":"<p><strong>Introduction and Context:</strong> At the start of the video (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>), David Haikney, Technical Product Director at Arm, welcomes viewers to the Microsoft Build session and introduces Arm Performix, a new performance analysis toolkit designed to help developers observe and accelerate workloads on the Cobalt platform. He highlights the close partnership between Microsoft and Arm, underlining that Performix was created in collaboration with Microsoft’s performance experts. David acknowledges that many viewers may be new to Cobalt and encourages them to explore available migration resources that can aid in their optimization journey (<a class=\"medius-aisummary\" data-starttime=\"00:00:33\">00:00:33</a>).</p>\n\n<p><strong>Understanding Performance Analysis:</strong> Before demonstrating the tool, David frames performance analysis as an investigative process (<a class=\"medius-aisummary\" data-starttime=\"00:00:44\">00:00:44</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:10\">00:01:10</a>). He compares it to a crime investigation that requires gathering evidence, forming hypotheses, and revising assumptions — an iterative and scientific method. Performix was built to simplify this complex process by unifying multiple analysis utilities. Effective performance analysis, he explains, begins by forming a broad understanding of the platform’s capabilities, then applying workloads to gather data before diving deeper into bottlenecks. This systematic approach ensures that experiments are consistent, measurable, and comparable across test runs (<a class=\"medius-aisummary\" data-starttime=\"00:01:26\">00:01:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>).</p>\n\n<p><strong>Basic Concepts and Setup:</strong> When introducing the Performix UI (<a class=\"medius-aisummary\" data-starttime=\"00:02:07\">00:02:07</a>), David explains that users can work through the interface or the command line. He introduces two foundational concepts: “targets” and “recipes.” Targets refer to the systems (in this case, Cobalt instances) where workloads will run, which can be configured easily using SSH credentials. Recipes are sets of experiments designed to analyze various aspects of system performance (<a class=\"medius-aisummary\" data-starttime=\"00:02:43\">00:02:43</a>). Some recipes are still experimental as the toolkit grows. To demonstrate, David runs a “system characterization” recipe that benchmarks memory bandwidth and cache performance to validate the Cobalt instance’s baseline capability (<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>).</p>\n\n<p><strong>Running Workload Analysis:</strong> The next demonstration focuses on observing system behavior under load using the “system utilization” recipe (<a class=\"medius-aisummary\" data-starttime=\"00:03:42\">00:03:42</a>). Users can launch a new workload, attach to an existing process, or analyze all system activity. A heat map visualization of a 96-core system is shown, illustrating that about two-thirds of the cores are actively used. This quick view helps analysts determine potential next steps, such as adjusting thread count or investigating idle cores for lock contention issues (<a class=\"medius-aisummary\" data-starttime=\"00:04:03\">00:04:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:30\">00:04:30</a>). David then demonstrates deeper analysis with the “code hotspots” recipe, which highlights time-intensive code regions across languages such as Java and .NET, allowing developers to focus optimization efforts precisely. He follows this by showing the “instruction mix” recipe and how runs can be compared between different optimizations — for example, using Arm NEON versus SVE instructions (<a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:33\">00:05:33</a>).</p>\n\n<p><strong>Deep Microarchitectural Insights:</strong> David continues by demonstrating the “CPU microarchitecture” recipe (<a class=\"medius-aisummary\" data-starttime=\"00:05:35\">00:05:35</a>), which leverages Arm’s top-down methodology to analyze how time is distributed across CPU functions. This analysis covers cache performance, pipeline behavior, and issues such as branch mispredictions. He emphasizes how Performix combines breadth-first and depth-first exploration — starting broadly to find trends and then drilling into specific causes — enabling a comprehensive investigation process (<a class=\"medius-aisummary\" data-starttime=\"00:05:55\">00:05:55</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:05\">00:06:05</a>).</p>\n\n<p><strong>Integrating AI and Final Summary:</strong> In the final segment (<a class=\"medius-aisummary\" data-starttime=\"00:06:10\">00:06:10</a>), David introduces how Performix integrates with an MCP server and large language models (LLMs) to transition from analysis to automatic insight generation. Within Visual Studio, profiling data, disassembly, and system context can be provided to the LLM, enabling dynamic feedback and confidence-based recommendations for performance improvement (<a class=\"medius-aisummary\" data-starttime=\"00:06:26\">00:06:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>). In closing, he summarizes that Performix’s recipe-based guided workflow helps users move methodically from broad system characterization to focused optimization. Now freely available for download and use, David invites developers to try the toolkit and share feedback (<a class=\"medius-aisummary\" data-starttime=\"00:07:00\">00:07:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:39\">00:07:39</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/96f9277a-f0b3-4e0c-8d91-55c4b9785084","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/96f9277a-f0b3-4e0c-8d91-55c4b9785084","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/96f9277a-f0b3-4e0c-8d91-55c4b9785084","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/96f9277a-f0b3-4e0c-8d91-55c4b9785084","speakerIds":["1768245294609build26-1776815051033001wbO0"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.638Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Arm"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/96f9277a-f0b3-4e0c-8d91-55c4b9785084","ogTitle":"Performance tuning on Cobalt with Arm Performix​","ogDescription":"Learn how to identify code hotspots, interpret performance data, and move from raw metrics to concrete improvements. The session also explores advanced command-line usage and shows how Arm Performix can be integrated into automated and agentic workflows for continuous performance analysis.​ By the end of the session, you’ll be able to apply Performix in real-world development pipelines to continuously monitor, diagnose, and optimize application performance.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776815051033001wbO0","RegistrantKey":"1768245294609build26-1776815051033001wbO0"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Automation","logicalValue":"Automation"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Local AI","logicalValue":"Local AI"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"ISV","logicalValue":"ISV"}]},{"@search.score":1.0,"documentId":"356087b0-4ffd-42b1-9065-ae54755ad41d","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Joydeep Mukherjee, Sam Shapiro","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774274079311001rAP6","langLocale":"en-US","sessionId":"1774274079311001rAP6","sessionInstanceId":"1774274079311001rAP6","sessionCode":"OD852","sessionCodeNormalized":"EN-US-OD852","title":"Powering enterprise-grade AI agents with Windows 365 for Agents","sortTitle":"powering enterprise-grade ai agents with windows 365 for agents","description":"Effective AI agents need GUI automation, legacy apps, or authenticated browser sessions with enterprise-grade security. Windows 365 for Agents meets this challenge by providing Cloud PCs that let AI agents—first or third party—operate inside full and secure computer environments. In this session, we’ll show how to enable agents to complete end-to-end tasks, integrate Windows 365 for Agents with Copilot Studio and custom frameworks, and manage Cloud PCs.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/7e2216bc-3d48-414f-8eb9-c647759df0be","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/7e2216bc-3d48-414f-8eb9-c647759df0be","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/7e2216bc-3d48-414f-8eb9-c647759df0be","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/7e2216bc-3d48-414f-8eb9-c647759df0be","speakerIds":["1768245294609build26-1766177772590001Xb4V","1768245294609build26-1780434812193001ekFj"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.635Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/7e2216bc-3d48-414f-8eb9-c647759df0be","ogTitle":"Powering enterprise-grade AI agents with Windows 365 for Agents","ogDescription":"Effective AI agents need GUI automation, legacy apps, or authenticated browser sessions with enterprise-grade security. Windows 365 for Agents meets this challenge by providing Cloud PCs that let AI agents—first or third party—operate inside full and secure computer environments. In this session, we’ll show how to enable agents to complete end-to-end tasks, integrate Windows 365 for Agents with Copilot Studio and custom frameworks, and manage Cloud PCs.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1766177772590001Xb4V","RegistrantKey":"1768245294609build26-1766177772590001Xb4V"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1780434812193001ekFj","RegistrantKey":"1768245294609build26-1780434812193001ekFj"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26/OD852","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows 365","logicalValue":"Windows 365"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows 365 for Agents","logicalValue":"Windows 365 for Agents"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"}]},{"@search.score":1.0,"documentId":"f8842f29-bf52-42ac-af3c-29c0c1021bdd","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Rakesh Krishnan","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693113266001Cpiu","langLocale":"en-US","sessionId":"1775693113266001Cpiu","sessionInstanceId":"1775693113266001Cpiu","sessionCode":"OD813","sessionCodeNormalized":"EN-US-OD813","title":"Powering modern data analytics in Fabric Data Warehouse","sortTitle":"powering modern data analytics in fabric data warehouse","description":"Developers are building interactive, AI‑driven analytics workflows, but the data layer often can’t keep up. Dashboards stall, pipelines miss SLAs, and agentic analytics break when queries are slow. In this session, learn about how a new engine in Fabric Data Warehouse delivers a faster, more predictable foundation accelerating reporting, pipelines, and agentic experiences without re‑architecting your stack.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/5e254e05-5173-4c48-8d44-cef28b28b5c5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/5e254e05-5173-4c48-8d44-cef28b28b5c5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/5e254e05-5173-4c48-8d44-cef28b28b5c5","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693113266001Cpiu_20260603024843614.avif","speakerIds":["1768245294609build26-1780418541035001Y0Yl"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693113266001Cpiu_20260603024843614.avif","ogTitle":"Powering modern data analytics in Fabric Data Warehouse","ogDescription":"Developers are building interactive, AI‑driven analytics workflows, but the data layer often can’t keep up. Dashboards stall, pipelines miss SLAs, and agentic analytics break when queries are slow. In this session, learn about how a new engine in Fabric Data Warehouse delivers a faster, more predictable foundation accelerating reporting, pipelines, and agentic experiences without re‑architecting your stack.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1780418541035001Y0Yl","RegistrantKey":"1768245294609build26-1780418541035001Y0Yl"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"20ee1483-aa7f-4bf8-a517-8244a1ce4bbf","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Amir Netz, Eren Orbey","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693037345001xUWz","langLocale":"en-US","sessionId":"1775693037345001xUWz","sessionInstanceId":"1775693037345001xUWz","sessionCode":"OD811","sessionCodeNormalized":"EN-US-OD811","title":"Powering the next AI frontier with a unified data platform","sortTitle":"powering the next ai frontier with a unified data platform","description":"AI agents are becoming part of the workforce, working side by side with employees. The challenge is making them as trusted and productive as people by giving them the same knowledge, context, and real‑time business understanding. In this session, explore the latest Microsoft Fabric capabilities making this possible, including Fabric IQ, and see how developers can build, ship, and scale AI apps on Fabric’s unified data foundation.","aiDescription":"<p><strong>Introduction and Fabric Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>, Amir Netz, CTO of Microsoft Fabric, introduces himself and explains how Microsoft Fabric was launched just over two years ago as a unified end-to-end data platform combining operational databases, SQL, data integration, analytics, data warehousing, and real-time intelligence into one ecosystem. He highlights key components like OneLake for unified data storage and Copilot as an integral capability spanning the stack. The adoption metrics are remarkable — Fabric is used by 35,000 customers, serving over 90% of Fortune 500 companies, and generating $2 billion in annual recurring revenue, growing 60% year-over-year. As he shifts focus at <a class=\"medius-aisummary\" data-starttime=\"00:01:39\">00:01:39</a>, Amir notes the rise of AI “agents” entering workplaces and poses a crucial question: how to make these agents as trustworthy and productive as human employees.</p>\n\n<p><strong>Introducing Microsoft IQ and Fabric IQ:</strong> From <a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a> onward, Amir outlines Microsoft’s answer — giving agents organizational context and knowledge comparable to employees. He introduces Microsoft IQ, comprising Work IQ, Foundry IQ, and Fabric IQ. Work IQ understands individuals and their work styles; Foundry IQ curates institutional knowledge; and Fabric IQ provides real-time business context. At <a class=\"medius-aisummary\" data-starttime=\"00:04:27\">00:04:27</a>, Amir explains that building Fabric IQ involves four steps: unifying data, harmonizing and processing it, curating semantic knowledge, and empowering AI agents. The first step centers on OneLake, Fabric’s secure open data foundation where all enterprise data is consolidated without duplication. By <a class=\"medius-aisummary\" data-starttime=\"00:07:09\">00:07:09</a>, he announces new data connectors for SAP, Oracle, SharePoint, OneDrive, and previews for AWS Glue and Azure Monitor, reinforcing a “zero-ETL” approach for complete data unification.</p>\n\n<p><strong>Performance, Agent Skills, and AI Functions:</strong> Starting at <a class=\"medius-aisummary\" data-starttime=\"00:08:08\">00:08:08</a>, Amir unveils GPU-accelerated query performance in the Fabric Data Warehouse, achieving up to 8x to 100x improvements under concurrent workloads — a breakthrough showcased in multiple demos. A technical walkthrough at <a class=\"medius-aisummary\" data-starttime=\"00:10:24\">00:10:24</a> shows how enabling “query acceleration” boosts insight retrieval speed dramatically. Later, Amir introduces “agent skills for Fabric” (<a class=\"medius-aisummary\" data-starttime=\"00:12:01\">00:12:01</a>), extending Fabric’s capabilities so AI agents can develop complete data solutions, ranging from ingestion to reporting. In a demonstration, Fabric’s CLI and GitHub Copilot skillfully generate ingestion pipelines, notebooks, scheduling pools, and even Power BI reports with near-instant startup times. This agent-driven approach shifts Fabric development from manual configuration to natural language-driven automation.</p>\n\n<p><strong>Semantics, Ontologies, and Planning in Fabric:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:17:20\">00:17:20</a>, the focus moves to creating business semantics so AI understands organizational context. Amir describes semantic models — the core of Power BI analytics — as the bridge between business language and stored data. He introduces ontologies (<a class=\"medius-aisummary\" data-starttime=\"00:18:58\">00:18:58</a>) to extend semantics beyond static metrics to encompass dynamic business processes, roles, policies, and actions. These ontologies model not only past and present states but also “should happen” scenarios, forming the foundation of operational intelligence. This naturally leads to the announcement of the general availability of “Planning in Fabric” at <a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>. The planning module lets organizations create forecasts, budgets, and scenarios natively within Fabric, fully integrated with OneLake and semantic models. A detailed demo illustrates how users manipulate assumptions interactively, lock scenarios, and collaborate with team members while maintaining complete data lineage and auditability.</p>\n\n<p><strong>Apps, Rayfin SDK, and Operational Agents:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:26:25\">00:26:25</a>, Amir introduces “Apps in Fabric,” allowing enterprises to build production-grade, data-connected applications through the new Rayfin SDK. The SDK scaffolds backends automatically, wiring authentication, SQL databases, and UI components with one command. Demos show Rayfin generating full delivery apps and extending into “agent-driven BI apps” (<a class=\"medius-aisummary\" data-starttime=\"00:31:00\">00:31:00</a>) that integrate operational and analytical workflows. Finally, the narration transitions at <a class=\"medius-aisummary\" data-starttime=\"00:33:05\">00:33:05</a> to “operations agents,” now generally available. Unlike chatbots, these agents monitor live signals, perform autonomous actions, and learn continuously. Through a hands-on example, an agent trained to reduce customer churn demonstrates natural language setup, semantic alignment through ontologies, integration with Microsoft Teams, and iterative learning via root-cause analysis — showing how virtual coworkers act proactively within business processes.</p>\n\n<p><strong>Conclusion and Vision:</strong> In the closing section at <a class=\"medius-aisummary\" data-starttime=\"00:37:12\">00:37:12</a>, Amir recaps the journey from data unification in OneLake through GPU-accelerated analytics, semantic modeling, planning, application development, and finally to autonomous agents. All these innovations, he emphasizes, are powered by Fabric IQ — the foundation that empowers both human employees and AI agents to operate with shared intelligence and context. As Microsoft expands Microsoft IQ across Work IQ, Foundry IQ, and Fabric IQ, Amir envisions a workforce of millions of intelligent agents collaborating productively alongside humans. He concludes by inviting viewers to explore these innovations and “have a great conference,” marking the session’s inspirational close at <a class=\"medius-aisummary\" data-starttime=\"00:38:11\">00:38:11</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4b833a99-c06b-4edf-b6a8-5063fc488c92","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4b833a99-c06b-4edf-b6a8-5063fc488c92","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4b833a99-c06b-4edf-b6a8-5063fc488c92","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4b833a99-c06b-4edf-b6a8-5063fc488c92","speakerIds":["1768245294609build26-1765523987882001Qt5i","1768245294609build26-1776283991549001CcBP"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.894Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4b833a99-c06b-4edf-b6a8-5063fc488c92","ogTitle":"Powering the next AI frontier with a unified data platform","ogDescription":"AI agents are becoming part of the workforce, working side by side with employees. The challenge is making them as trusted and productive as people by giving them the same knowledge, context, and real‑time business understanding. In this session, explore the latest Microsoft Fabric capabilities making this possible, including Fabric IQ, and see how developers can build, ship, and scale AI apps on Fabric’s unified data foundation.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) Digital","SpeakerId":"1768245294609build26-1765523987882001Qt5i","RegistrantKey":"1768245294609build26-1765523987882001Qt5i"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776283991549001CcBP","RegistrantKey":"1768245294609build26-1776283991549001CcBP"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Linux","logicalValue":"Linux"},{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"7a633d6b-b496-4fa8-8082-34162868bb6b","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Arindam Chatterjee, Tessa Kloster, Anshul Sharma","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693251665001bn9w","langLocale":"en-US","sessionId":"1775693251665001bn9w","sessionInstanceId":"1775693251665001bn9w","sessionCode":"OD819","sessionCodeNormalized":"EN-US-OD819","title":"Real-Time Intelligence: Building\u00A0event-driven\u00A0AI\u00A0apps\u00A0and\u00A0agents","sortTitle":"real-time intelligence: building\u00A0event-driven\u00A0ai\u00A0apps\u00A0and\u00A0agents","description":"Join this session to see how Microsoft Fabric Real‑Time Intelligence helps developers build event‑driven AI apps and autonomous agents that respond to live data in seconds. Real‑Time Intelligence unifies streaming ingestion, real‑time analytics, and actioning in a single governed experience—enabling teams to move seamlessly from signal to insight to action without complex pipelines.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/af42f4ba-68b0-4022-a283-f6db86482755","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/af42f4ba-68b0-4022-a283-f6db86482755","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/af42f4ba-68b0-4022-a283-f6db86482755","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/af42f4ba-68b0-4022-a283-f6db86482755","speakerIds":["1768245294609build26-1777498254683001UmWt","1768245294609build26-1765361984218001ZKhA","1768245294609build26-1770501028148001j5q7"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.897Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/af42f4ba-68b0-4022-a283-f6db86482755","ogTitle":"Real-Time Intelligence: Building\u00A0event-driven\u00A0AI\u00A0apps\u00A0and\u00A0agents","ogDescription":"Join this session to see how Microsoft Fabric Real‑Time Intelligence helps developers build event‑driven AI apps and autonomous agents that respond to live data in seconds. Real‑Time Intelligence unifies streaming ingestion, real‑time analytics, and actioning in a single governed experience—enabling teams to move seamlessly from signal to insight to action without complex pipelines.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1765361984218001ZKhA","RegistrantKey":"1768245294609build26-1765361984218001ZKhA"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777498254683001UmWt","RegistrantKey":"1768245294609build26-1777498254683001UmWt"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1770501028148001j5q7","RegistrantKey":"1768245294609build26-1770501028148001j5q7"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aks.ms/build26/OD819","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"add31060-d22d-4369-a425-9eb6658e419d","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Carlos Camacho","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396747350001xik8","langLocale":"en-US","sessionId":"1774396747350001xik8","sessionInstanceId":"1774396747350001xik8","sessionCode":"ODSP915","sessionCodeNormalized":"EN-US-ODSP915","title":"Resolve support cases with multi-agent workflows","sortTitle":"resolve support cases with multi-agent workflows","description":"See how a multi-agent system diagnoses and resolves support cases across partner ecosystems. This demo showcases the Agentic Partners Integration AI quickstarts, built on Azure and Red Hat OpenShift AI (ARO) with MCP and open-source components. Learn how agents share context, orchestrate actions, and automate resolution. Leave with a reference architecture and patterns to build your own agentic workflows.","aiDescription":"<p><strong>Introduction and Objective:</strong> The video begins with Carlos and Sharon from Red Hat greeting viewers <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a> and explaining that they will demonstrate how to deploy an MCP server from the MCP catalog and use it in a multi-agent application that diagnoses and resolves support cases across partner ecosystems. They emphasize that the session will be brief, fully live, and focused on facilitating support resolution across various partners within one ecosystem <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>. The goal is to show an open-source, multi-agent system that enables automated workflows to diagnose and fix problems efficiently.</p>\n\n<p><strong>System Architecture Overview:</strong> The presenters describe the layers of the system starting with the user access layer <a class=\"medius-aisummary\" data-starttime=\"00:00:52\">00:00:52</a>, a PatternFly-based web app that allows interaction through a chat interface. They then explain the orchestration and security layer <a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>, which manages authentication, authorization, and auditing. Next, Carlos details the routing agent <a class=\"medius-aisummary\" data-starttime=\"00:01:17\">00:01:17</a>, responsible for directing user requests to the appropriate troubleshooting agents. The communication between agents is facilitated by A2A, and all agents are developed using ADK for modular integration within the ecosystem <a class=\"medius-aisummary\" data-starttime=\"00:01:31\">00:01:31</a>.</p>\n\n<p><strong>Deploying the Azure MCP Server:</strong> Sharon explains the process of using the OpenShift AI MCP Catalog <a class=\"medius-aisummary\" data-starttime=\"00:01:41\">00:01:41</a> to deploy a pre-built MCP server. The catalog offers ready-to-deploy servers that can be configured and launched with minimal effort. In this demo, they choose to deploy the Azure MCP server to their Red Hat OpenShift cluster <a class=\"medius-aisummary\" data-starttime=\"00:01:51\">00:01:51</a>. She mentions that users only need to specify deployment name, project, and Azure tenant details. The catalog automatically manages container images, configurations, and authentication through managed identity <a class=\"medius-aisummary\" data-starttime=\"00:02:06\">00:02:06</a>. Importantly, the deployment requires no custom coding—just selecting and deploying from the catalog and consuming it like any other workload <a class=\"medius-aisummary\" data-starttime=\"00:02:17\">00:02:17</a>. Once configured, users click “Deploy the Azure MCP server” and wait for the operator to complete the process <a class=\"medius-aisummary\" data-starttime=\"00:02:27\">00:02:27</a>.</p>\n\n<p><strong>Demonstration of Multi-Agent Workflow:</strong> After deployment, Carlos logs into the quickstart environment <a class=\"medius-aisummary\" data-starttime=\"00:02:39\">00:02:39</a> using a user account with access to several agents, including the Azure support agent. He begins by asking questions related to Azure services, such as details on the Well-Architected Framework for AKS <a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. The routing agent identifies the intent, delegates execution to the Azure support agent, and displays the output generated by underlying MCP tools <a class=\"medius-aisummary\" data-starttime=\"00:03:08\">00:03:08</a>. Each query triggers two phases: intent detection and MCP tool execution <a class=\"medius-aisummary\" data-starttime=\"00:03:20\">00:03:20</a>. He further queries best practices for Azure web application development <a class=\"medius-aisummary\" data-starttime=\"00:03:30\">00:03:30</a>, showing similar routing and execution flow that produces a markdown file answer <a class=\"medius-aisummary\" data-starttime=\"00:03:51\">00:03:51</a>.</p>\n\n<p><strong>Auditing, Security, and Access Control:</strong> The video highlights built-in auditing capabilities <a class=\"medius-aisummary\" data-starttime=\"00:04:16\">00:04:16</a>. All application events are logged using OpenTelemetry, giving visibility into user activity and request permissions <a class=\"medius-aisummary\" data-starttime=\"00:04:21\">00:04:21</a>. The audit trail records whether specific requests were allowed or denied and helps trace group-based access rights. Carlos illustrates this by switching to a user named Josh <a class=\"medius-aisummary\" data-starttime=\"00:04:48\">00:04:48</a>, who lacks permission to any agents. When Josh attempts to submit a request, the routing agent denies it <a class=\"medius-aisummary\" data-starttime=\"00:04:55\">00:04:55</a>, and the event appears in the audit logs and traces <a class=\"medius-aisummary\" data-starttime=\"00:05:04\">00:05:04</a>. This confirms that access control and auditing are active and reliable.</p>\n\n<p><strong>Conclusion:</strong> To conclude, the presenters note that users without assigned departments or agent access, such as Josh <a class=\"medius-aisummary\" data-starttime=\"00:05:13\">00:05:13</a>, will not have requests processed, validating the system's layered security design. They end the demo by summarizing the simplicity and robustness of deploying MCP servers through the Red Hat AI MCP Catalog and integrating them into secure, multi-agent workflows for support resolution <a class=\"medius-aisummary\" data-starttime=\"00:05:20\">00:05:20</a>. The session closes with a brief thank-you to the audience <a class=\"medius-aisummary\" data-starttime=\"00:05:22\">00:05:22</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/95943165-d73d-4156-8561-753ad666eb99","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/95943165-d73d-4156-8561-753ad666eb99","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/95943165-d73d-4156-8561-753ad666eb99","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/95943165-d73d-4156-8561-753ad666eb99","speakerIds":["1768245294609build26-1777508179227001evcF"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.64Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Red Hat"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/95943165-d73d-4156-8561-753ad666eb99","ogTitle":"Resolve support cases with multi-agent workflows","ogDescription":"See how a multi-agent system diagnoses and resolves support cases across partner ecosystems. This demo showcases the Agentic Partners Integration AI quickstarts, built on Azure and Red Hat OpenShift AI (ARO) with MCP and open-source components. Learn how agents share context, orchestrate actions, and automate resolution. Leave with a reference architecture and patterns to build your own agentic workflows.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777508179227001evcF","RegistrantKey":"1768245294609build26-1777508179227001evcF"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Vector Embeddings","logicalValue":"Vector Embeddings"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"OSS","logicalValue":"OSS"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"ISV","logicalValue":"ISV"},{"displayValue":"Open Ecosystem","logicalValue":"Open Ecosystem"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"}]},{"@search.score":1.0,"documentId":"bf2bf945-f51e-4fc7-aa85-ece063054d73","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Daniel Arrizza, Katarina Stanley","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396692596001IXfu","langLocale":"en-US","sessionId":"1774396692596001IXfu","sessionInstanceId":"1774396692596001IXfu","sessionCode":"ODSP914","sessionCodeNormalized":"EN-US-ODSP914","title":"Run AI at scale with Ray + Kubernetes using Anyscale on Azure","sortTitle":"run ai at scale with ray + kubernetes using anyscale on azure","description":"In this session see how to build multimodal data pipelines, train and fine-tune models, and deploy them as inference services, all without managing infrastructure. Walk through the end-to-end architecture and run a live demo of building, training, and serving AI inside your Azure subscription. Anyscale on Azure is a production-ready AI platform built on a scalable stack of Ray and Azure Kubernetes Service that abstracts the complexity of distributed compute behind a Python-native interface.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/a7be3cd5-4bb1-448a-b4fa-ee62f93aad47","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/a7be3cd5-4bb1-448a-b4fa-ee62f93aad47","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/a7be3cd5-4bb1-448a-b4fa-ee62f93aad47","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/a7be3cd5-4bb1-448a-b4fa-ee62f93aad47","speakerIds":["1768245294609build26-1777486041744001HXer","1768245294609build26-1777331816832001zQaj"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.641Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Anyscale"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/a7be3cd5-4bb1-448a-b4fa-ee62f93aad47","ogTitle":"Run AI at scale with Ray + Kubernetes using Anyscale on Azure","ogDescription":"In this session see how to build multimodal data pipelines, train and fine-tune models, and deploy them as inference services, all without managing infrastructure. Walk through the end-to-end architecture and run a live demo of building, training, and serving AI inside your Azure subscription. Anyscale on Azure is a production-ready AI platform built on a scalable stack of Ray and Azure Kubernetes Service that abstracts the complexity of distributed compute behind a Python-native interface.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Speaker Digital","SpeakerId":"1768245294609build26-1777331816832001zQaj","RegistrantKey":"1768245294609build26-1777331816832001zQaj"},{"SpeakerType":"Owner, Session Owner, Speaker Digital","SpeakerId":"1768245294609build26-1777486041744001HXer","RegistrantKey":"1768245294609build26-1777486041744001HXer"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Azure","logicalValue":"Azure"},{"displayValue":"Cost Management","logicalValue":"Cost Management"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Compute","logicalValue":"Compute"},{"displayValue":"Platform","logicalValue":"Platform"},{"displayValue":"Vector Embeddings","logicalValue":"Vector Embeddings"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Azure Kubernetes Service (AKS)​​","logicalValue":"Azure Kubernetes Service (AKS)​​"},{"displayValue":"Python","logicalValue":"Python"},{"displayValue":"PyTorch","logicalValue":"PyTorch"},{"displayValue":"Microsoft for Startups","logicalValue":"Microsoft for Startups"},{"displayValue":"AKS","logicalValue":"AKS"},{"displayValue":"Data","logicalValue":"Data"},{"displayValue":"Scaling","logicalValue":"Scaling"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"},{"displayValue":"Development pipeline","logicalValue":"Development pipeline"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"86c49d12-0827-4f99-8b9f-eb26aaab5bd7","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Natan Yellin","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776206958046001Zryy","langLocale":"en-US","sessionId":"1776206958046001Zryy","sessionInstanceId":"1776206958046001Zryy","sessionCode":"ODSP928","sessionCodeNormalized":"EN-US-ODSP928","title":"Run AI SREs without burning token budgets","sortTitle":"run ai sres without burning token budgets","description":"Most AI SREs have an underlying cost per investigation of about $2 per alert. This makes it difficult for most enterprises to use AI SREs as a first-line solution for triaging alerts, based on token cost. In this session, we review the math and optimizations and show practival ways to bring the LLM cost of AI SREs into a manageable range, where running them on every alert becomes viable.","aiDescription":"<p><strong>Introduction and Context:</strong> At the beginning of the video (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>), Natan Yellin introduces himself and his topic: building affordable and scalable SRE (Site Reliability Engineering) agents for production alerts. He explains that while many companies are trying to build SRE agents, the economic feasibility is a major challenge. As the CEO and co-founder of Robusta (<a class=\"medius-aisummary\" data-starttime=\"00:00:25\">00:00:25</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>), Yellin draws on extensive experience delivering these agents to clients ranging from Fortune 500 corporations to small startups. He sets the stage by committing to share practical lessons learned from deploying SRE agents in varied operational environments.</p>\n\n<p><strong>Understanding the Cost Problem:</strong> Yellin breaks down the cost calculation behind AI-driven SRE agents, emphasizing that naïve implementations are prohibitively expensive (<a class=\"medius-aisummary\" data-starttime=\"00:00:43\">00:00:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:56\">00:00:56</a>). He illustrates with an example where a typical large organization could spend $730,000 yearly if each investigation costs around $2 and handles approximately 1,000 alerts daily (<a class=\"medius-aisummary\" data-starttime=\"00:00:59\">00:00:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:07\">00:01:07</a>). He notes that this figure is often higher than manual alert triage performed by human operators, making it economically unviable. Yellin details how investigation prices vary—typically between 50 cents and $5 per alert (<a class=\"medius-aisummary\" data-starttime=\"00:01:49\">00:01:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:21\">00:02:21</a>)—and how vendors amplify these costs with markups, sometimes charging between $15 and $25 per alert. The foundational cost driver, he explains, comes from input tokens consumed by AI models when analyzing logs, metrics, and related data rather than output tokens.</p>\n\n<p><strong>Exploring Model and Configuration Optimizations:</strong> Transitioning from cost analysis, Yellin examines optimization opportunities by adjusting model selection and architecture (<a class=\"medius-aisummary\" data-starttime=\"00:03:04\">00:03:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:07\">00:03:07</a>). He explores how using cheaper large language models, such as DeepSeek V4 (<a class=\"medius-aisummary\" data-starttime=\"00:03:45\">00:03:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:50\">00:03:50</a>), could theoretically reduce expenses compared to Opus family models. However, he cautions that cheaper models may yield lower accuracy and often consume more tokens overall because they retrieve incorrect data first and reattempt queries (<a class=\"medius-aisummary\" data-starttime=\"00:04:26\">00:04:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:47\">00:04:47</a>). Despite this potential inefficiency, he acknowledges that switching models remains a baseline optimization that teams can easily experiment with, although Robusta prefers sticking with higher-performing Opus models due to better real-world results and other in-place cost management techniques.</p>\n\n<p><strong>Reducing Investigation Redundancy with Context and Grouping:</strong> Yellin moves into advanced strategies for cost reduction, highlighting contextual memory and alert grouping (<a class=\"medius-aisummary\" data-starttime=\"00:04:51\">00:04:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:56\">00:05:56</a>). He explains that without memory or reusable context, SRE agents must rediscover environment-specific details and observability configurations for every alert. Enabling context persistence or auto-learning can cut costs by 20–30%, as agents avoid redundant queries (<a class=\"medius-aisummary\" data-starttime=\"00:05:38\">00:05:38</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:53\">00:05:53</a>). The largest optimization, however, comes from LLM-native grouping of alerts, where similar or duplicate alerts—often hundreds triggered during a single outage—are investigated collectively (<a class=\"medius-aisummary\" data-starttime=\"00:05:59\">00:05:59</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:20\">00:06:20</a>). Yellin contrasts this AI-driven grouping with traditional deterministic rules, which fail to accurately cluster alerts across systems. Robusta’s methodology lets the LLM dynamically create and persist grouping rules (<a class=\"medius-aisummary\" data-starttime=\"00:07:04\">00:07:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:19\">00:07:19</a>), optimizing both accuracy and efficiency.</p>\n\n<p><strong>Technical Optimizations and Performance Gains:</strong> Adding more technical depth, Yellin describes how caching and context reuse further reduce costs (<a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:54\">00:07:54</a>). When an investigation is triggered repeatedly for related alerts, the system can reuse the same context window if it remains active within Anthropic’s five-minute cache cycle. This method substantially lowers computation fees compared with regenerating a fresh context window each time. These combined optimizations—intelligent grouping, caching, and context reuse—collectively enable investigations to scale across all alerts efficiently without compromising accuracy or triggering unnecessary repeats (<a class=\"medius-aisummary\" data-starttime=\"00:07:57\">00:07:57</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:05\">00:08:05</a>).</p>\n\n<p><strong>Conclusion and Invitation:</strong> Yellin concludes by reaffirming that AI-based SRE agents can achieve remarkable accuracy but often fail to deliver cost-effectiveness unless carefully optimized (<a class=\"medius-aisummary\" data-starttime=\"00:08:51\">00:08:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:05\">00:09:05</a>). Once cost per investigation falls sufficiently, organizations can confidently deploy SRE agents for every alert, unlocking proactive performance improvements and automated escalation (<a class=\"medius-aisummary\" data-starttime=\"00:08:13\">00:08:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:46\">00:08:46</a>). He closes by inviting viewers to reach out via email for deeper discussions and implementation guidance, emphasizing his willingness to share applied insights drawn from extensive field experience in automated reliability operations (<a class=\"medius-aisummary\" data-starttime=\"00:09:07\">00:09:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:13\">00:09:13</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/87ba3ffc-a093-4698-849e-3a79fc894ccd","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/87ba3ffc-a093-4698-849e-3a79fc894ccd","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/87ba3ffc-a093-4698-849e-3a79fc894ccd","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/87ba3ffc-a093-4698-849e-3a79fc894ccd","speakerIds":["1768245294609build26-1777193724498001kkv2"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.902Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Robusta Dev"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/87ba3ffc-a093-4698-849e-3a79fc894ccd","ogTitle":"Run AI SREs without burning token budgets","ogDescription":"Most AI SREs have an underlying cost per investigation of about $2 per alert. This makes it difficult for most enterprises to use AI SREs as a first-line solution for triaging alerts, based on token cost. In this session, we review the math and optimizations and show practival ways to bring the LLM cost of AI SREs into a manageable range, where running them on every alert becomes viable.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Owner, Speaker Digital","SpeakerId":"1768245294609build26-1777193724498001kkv2","RegistrantKey":"1768245294609build26-1777193724498001kkv2"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Outages","logicalValue":"Outages"},{"displayValue":"Platform Engineering","logicalValue":"Platform Engineering"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Platform","logicalValue":"Platform"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Production Systems","logicalValue":"Production Systems"}]},{"@search.score":1.0,"documentId":"c34eb9ca-2960-4534-a008-a56422d94897","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Arpitha Dhanapathi","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776897866893001DxAZ","langLocale":"en-US","sessionId":"1776897866893001DxAZ","sessionInstanceId":"1776897866893001DxAZ","sessionCode":"OD829","sessionCodeNormalized":"EN-US-OD829","title":"Safeguard data across you apps and agents development cycle","sortTitle":"safeguard data across you apps and agents development cycle","description":"Developers ship fast—but AI apps can overshare data and create compliance risks. Learn why data security and compliance must be built in from day one, and how to align with enterprise policies without reinventing controls. We’ll walk through common leakage paths, key guardrails, and how integrating app and agent development with Microsoft Purview can help your team move from prototype to production with confidence.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/05a745d2-c8f6-4437-9369-3e3b51f2ecc7","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/05a745d2-c8f6-4437-9369-3e3b51f2ecc7","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/05a745d2-c8f6-4437-9369-3e3b51f2ecc7","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/05a745d2-c8f6-4437-9369-3e3b51f2ecc7","speakerIds":["1768245294609build26-1768337469692001nSaX"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.387Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/05a745d2-c8f6-4437-9369-3e3b51f2ecc7","ogTitle":"Safeguard data across you apps and agents development cycle","ogDescription":"Developers ship fast—but AI apps can overshare data and create compliance risks. Learn why data security and compliance must be built in from day one, and how to align with enterprise policies without reinventing controls. We’ll walk through common leakage paths, key guardrails, and how integrating app and agent development with Microsoft Purview can help your team move from prototype to production with confidence.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1768337469692001nSaX","RegistrantKey":"1768245294609build26-1768337469692001nSaX"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"}]},{"@search.score":1.0,"documentId":"273485f0-2a44-4cb1-aa46-81ec41d16edd","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799393343001DezL","langLocale":"en-US","sessionId":"1776799393343001DezL","sessionInstanceId":"1776799393343001DezL","sessionCode":"OD824","sessionCodeNormalized":"EN-US-OD824","title":"Scalable Applications Without Polyglot tax: Azure SQL Hyperscale","sortTitle":"scalable applications without polyglot tax: azure sql hyperscale","description":"Modern application architectures have drifted toward polyglot persistence, adding complexity, data movement, and operational overhead. Azure SQL Database Hyperscale takes a different approach: a single, multi model system that brings relational, transactional, and emerging workloads together within SQL. Learn how Hyperscale redefines performance, elasticity, and operations. See why consolidating on a single, multi-model SQL platform is the simpler, more scalable path for modern applications","aiDescription":"<p><strong>Introduction and the Problem of Polyglot Tax:</strong> The talk begins with Aditya, a product manager from Azure SQL, greeting the audience and setting the stage for a discussion on building scalable applications without the “polyglot tax” <a class=\"medius-aisummary\" data-starttime=\"00:00:03\">00:00:03</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:17\">00:00:17</a>. He explains that polyglot tax refers to the cost—both technical and operational—of using multiple specialized databases within an application. As developers move from quick prototypes to production-scale systems, they often add vector databases for semantic search, graph databases for relationships, and analytics engines to extract insights. However, each additional database introduces complexity, data synchronization issues, and potential points of failure. Aditya stresses that this architectural tax manifests as maintenance overhead, network latency, multiple security systems, and cognitive load on developers <a class=\"medius-aisummary\" data-starttime=\"00:01:00\">00:01:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:00\">00:03:00</a>. The core message is that Microsoft SQL can eliminate these inefficiencies by integrating all necessary capabilities within one engine.</p>\n\n<p><strong>Demonstrating Polyglot Complexity with a Fraud Detection Example:</strong> To illustrate the challenge, Aditya describes how a team might build a real-time fraud detection system <a class=\"medius-aisummary\" data-starttime=\"00:02:10\">00:02:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:40\">00:03:40</a>. Such a system often requires separate databases: relational for transaction history, JSON/document stores for device fingerprints, graph databases for customer connections, vector stores for similarity matching, and analytic databases for statistical baselines. He highlights the ensuing problems—data consistency issues, synchronization lags, increased latency, and the costly need to maintain different backups and security policies. This polyglot setup multiplies both operational costs and developer burden, while also leading to unreliable outcomes for AI and agent-based systems that draw on inconsistent data. The goal, Aditya emphasizes, is to consolidate these workloads into Microsoft SQL to avoid these inefficiencies and improve system reliability <a class=\"medius-aisummary\" data-starttime=\"00:04:00\">00:04:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:10\">00:05:10</a>.</p>\n\n<p><strong>Microsoft SQL Core Engine Capabilities – JSON, Graph, and Vector Integration:</strong> Transitioning into “Act 1,” Aditya outlines how the Microsoft SQL Core Engine itself helps avoid polyglot tax. He introduces native JSON data types and JSON indexing, which allow flexible document-style storage directly within SQL tables <a class=\"medius-aisummary\" data-starttime=\"00:06:00\">00:06:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:30\">00:07:30</a>. Unlike earlier implementations, this is a pre-parsed binary type that enhances performance and compresses data by 30–50%. SQL now supports indexing at specific JSON paths for faster lookups, rivaling leading document databases. He continues by describing inbuilt graph processing support that enables relationship modeling and pattern matching natively within the SQL query optimizer <a class=\"medius-aisummary\" data-starttime=\"00:08:10\">00:08:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:09:00\">00:09:00</a>. Aditya then details vector data support—a newer addition allowing vector storage and approximate nearest-neighbor searches via Microsoft’s DiskANN index, derived from Microsoft Research <a class=\"medius-aisummary\" data-starttime=\"00:09:20\">00:09:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:55\">00:10:55</a>. With these integrations—JSON, Graph, Vector, and clustered columnstore for analytics—Microsoft SQL can now handle transactional, analytical, and AI-driven workloads in a single ACID-compliant engine.</p>\n\n<p><strong>Unified Querying and Transaction Boundaries:</strong> Aditya demonstrates how these integrated features converge in practice <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>. For example, a developer can join relational order data, JSON device fingerprints, vector similarity results, and graph relationships across a single stored procedure—executed under one consistent transaction scope. This eliminates the risk of partial writes or inconsistencies that plague heterogeneous systems. He notes that such ACID guarantees are particularly critical for agents and AI-driven processes that require determinism and rollback safety <a class=\"medius-aisummary\" data-starttime=\"00:14:00\">00:14:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:00\">00:15:00</a>. With all storage and querying consolidated, Microsoft SQL ensures a single source of truth across multiple data modalities while reducing developer effort to one query language and tooling environment.</p>\n\n<p><strong>Azure SQL Hyperscale – Architecture, Scalability, and Reliability:</strong> In “Act 2,” the discussion shifts to the Azure SQL Hyperscale service, which extends this core engine into a scalable cloud architecture <a class=\"medius-aisummary\" data-starttime=\"00:15:10\">00:15:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:18:00\">00:18:00</a>. Aditya calls it the “Peace of Mind” database because it virtually eliminates capacity worries—storage automatically scales up to 128 TB, with separate dedicated services for logging, page management, and backups. Built-in snapshot-based backups avoid compute overhead, increasing efficiency and reducing downtime. Hyperscale also introduces named replicas for read scalability—each with independent cache and compute resources—allowing developers to separate OLTP traffic from analytical or agentic workloads. The platform inherently supports high availability, geo-replication, and serverless scaling that adapts compute power according to load <a class=\"medius-aisummary\" data-starttime=\"00:19:00\">00:19:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:00\">00:21:00</a>. Additionally, built-in features such as Always Encrypted, dynamic data masking, and tamper-evident ledger functions enhance data security, while integration with the Data API Builder and Microsoft Copilot (MCP server) enables natural language interaction with SQL data.</p>\n\n<p><strong>Conclusion – Economics and Strategic Takeaways:</strong> Aditya concludes by emphasizing the economic and architectural advantages of adopting Microsoft SQL and Azure SQL Hyperscale <a class=\"medius-aisummary\" data-starttime=\"00:21:31\">00:21:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:23:20\">00:23:20</a>. By avoiding specialized polyglot systems, organizations eliminate redundant data pipelines, reduce compute and storage costs, and streamline developer operations under a single licensing-free model. Hyperscale’s capability to dynamically scale compute and storage minimizes over-provisioning while achieving up to 68% better performance than comparable systems. Aditya closes with a pragmatic note: while some extreme workloads may still need specialized databases, for the majority of applications Microsoft SQL delivers a unified, scalable, secure, and cost-effective foundation <a class=\"medius-aisummary\" data-starttime=\"00:23:40\">00:23:40</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:19\">00:24:19</a>. He ends by encouraging developers to “avoid the polyglot tax, ship with confidence, scale efficiently, and sleep peacefully at night.”</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4610847f-3c55-4710-af17-90eaee6d192e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4610847f-3c55-4710-af17-90eaee6d192e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4610847f-3c55-4710-af17-90eaee6d192e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4610847f-3c55-4710-af17-90eaee6d192e","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4610847f-3c55-4710-af17-90eaee6d192e","ogTitle":"Scalable Applications Without Polyglot tax: Azure SQL Hyperscale","ogDescription":"Modern application architectures have drifted toward polyglot persistence, adding complexity, data movement, and operational overhead. Azure SQL Database Hyperscale takes a different approach: a single, multi model system that brings relational, transactional, and emerging workloads together within SQL. Learn how Hyperscale redefines performance, elasticity, and operations. See why consolidating on a single, multi-model SQL platform is the simpler, more scalable path for modern applications","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure SQL Hyperscale","logicalValue":"Azure SQL Hyperscale"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"906e2732-c7e0-45bc-98ce-5123de39f357","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Kim Manis","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693187955001xtRb","langLocale":"en-US","sessionId":"1775693187955001xtRb","sessionInstanceId":"1775693187955001xtRb","sessionCode":"OD816","sessionCodeNormalized":"EN-US-OD816","title":"Securing, scaling, and sustaining your data estate in Microsoft Fabric","sortTitle":"securing, scaling, and sustaining your data estate in microsoft fabric","description":"As GenAI adoption accelerates, developers and admins alike need control over governance, capacity, and security. In this session, we’ll walk through how Fabric helps you monitor your workloads, enforce governance policies, and manage capacity and access with precision. We’ll also show how developers can access their data and governance from the tools they use everyday through Fabric MCP, Agent Skills in Fabric, the Fabric CLI.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/176dd97c-df23-490c-8ef0-b8ae125c5792","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/176dd97c-df23-490c-8ef0-b8ae125c5792","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/176dd97c-df23-490c-8ef0-b8ae125c5792","onDemandThumbnail":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693187955001xtRb_20260603024851636.avif","speakerIds":["1768245294609build26-1780424923435001IyeK"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://mediastream.microsoft.com/Build2026/V1/itse/1775693187955001xtRb_20260603024851636.avif","ogTitle":"Securing, scaling, and sustaining your data estate in Microsoft Fabric","ogDescription":"As GenAI adoption accelerates, developers and admins alike need control over governance, capacity, and security. In this session, we’ll walk through how Fabric helps you monitor your workloads, enforce governance policies, and manage capacity and access with precision. We’ll also show how developers can access their data and governance from the tools they use everyday through Fabric MCP, Agent Skills in Fabric, the Fabric CLI.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Executive (CVP+) Digital","SpeakerId":"1768245294609build26-1780424923435001IyeK","RegistrantKey":"1768245294609build26-1780424923435001IyeK"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"850d0fb3-7025-4528-b5d8-69ed478e5fba","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Klorida Miraj, Nazmus Sakib","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774274079471001r1BH","langLocale":"en-US","sessionId":"1774274079471001r1BH","sessionInstanceId":"1774274079471001r1BH","sessionCode":"OD853","sessionCodeNormalized":"EN-US-OD853","title":"Security for AI and Agent Augmented Coding on Windows","sortTitle":"security for ai and agent augmented coding on windows","description":"AI agents create new opportunities for Windows developers to help users do more with less effort. This session explores how agent development on Windows benefits from built‑in security, observability, and manageability at the OS level. Using real examples like sandbox in GitHub CLI , we show how de‑privileged execution and clear boundaries let developers safely move agents from experiments to production without sacrificing trust or velocity.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/eaa729a7-8374-4730-91ec-695dd408faeb","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/eaa729a7-8374-4730-91ec-695dd408faeb","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/eaa729a7-8374-4730-91ec-695dd408faeb","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/eaa729a7-8374-4730-91ec-695dd408faeb","speakerIds":["1768245294609build26-1774285003250001pqCF","1768245294609build26-1774285003098001pwv1"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.635Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/eaa729a7-8374-4730-91ec-695dd408faeb","ogTitle":"Security for AI and Agent Augmented Coding on Windows","ogDescription":"AI agents create new opportunities for Windows developers to help users do more with less effort. This session explores how agent development on Windows benefits from built‑in security, observability, and manageability at the OS level. Using real examples like sandbox in GitHub CLI , we show how de‑privileged execution and clear boundaries let developers safely move agents from experiments to production without sacrificing trust or velocity.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1774285003098001pwv1","RegistrantKey":"1768245294609build26-1774285003098001pwv1"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1774285003250001pqCF","RegistrantKey":"1768245294609build26-1774285003250001pqCF"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Containment","logicalValue":"Containment"}]},{"@search.score":1.0,"documentId":"e3b8c35b-2aab-422e-b653-e39b7c9d6934","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"James Brotsos","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776898293448001fR7M","langLocale":"en-US","sessionId":"1776898293448001fR7M","sessionInstanceId":"1776898293448001fR7M","sessionCode":"OD841","sessionCodeNormalized":"EN-US-OD841","title":"Ship fast, stay secure: from code to runtime","sortTitle":"ship fast, stay secure: from code to runtime","description":"You write the code. You own the pipeline. Now security is yours too — but it doesn't have to slow you down. See how Defender for Cloud and GitHub Advanced Security catch vulnerabilities where you already work: your CLI, your repo, your pull request, your cloud. No workflow changes required.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4851ca31-b4e6-4830-9a88-d47365be3bc2","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4851ca31-b4e6-4830-9a88-d47365be3bc2","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4851ca31-b4e6-4830-9a88-d47365be3bc2","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4851ca31-b4e6-4830-9a88-d47365be3bc2","speakerIds":["1768245294609build26-1772736447689001HZZe"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.388Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4851ca31-b4e6-4830-9a88-d47365be3bc2","ogTitle":"Ship fast, stay secure: from code to runtime","ogDescription":"You write the code. You own the pipeline. Now security is yours too — but it doesn't have to slow you down. See how Defender for Cloud and GitHub Advanced Security catch vulnerabilities where you already work: your CLI, your repo, your pull request, your cloud. No workflow changes required.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Responsible AI","logicalValue":"Responsible AI"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1772736447689001HZZe","RegistrantKey":"1768245294609build26-1772736447689001HZZe"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Security","logicalValue":"Security"},{"displayValue":"GitHub Advanced Security","logicalValue":"GitHub Advanced Security"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"DevSecOps","logicalValue":"DevSecOps"}]},{"@search.score":1.0,"documentId":"eec1acd2-4857-48a3-9f5d-e60f9d20ff25","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776442375747001ZXvp","langLocale":"en-US","sessionId":"1776442375747001ZXvp","sessionInstanceId":"1776442375747001ZXvp","sessionCode":"ODSP932","sessionCodeNormalized":"EN-US-ODSP932","title":"Ship safer code from your IDE with Aikido","sortTitle":"ship safer code from your ide with aikido","description":"Learn how you can integrate security scanning directly into the IDE with Aikido. Using Aikido’s extension, you can ensure that you are shipping less vulnerabilities with your PRs, are not introducing secrets into your commits, and even block malware from making it onto your machine. Best of all, it's extremely lightweight and doesn't slow you down.","aiDescription":"<p><strong>Introduction and Purpose:</strong> Simon from Aikido opens the video by introducing himself and explaining the topic of the demonstration — the Aikido plugin for VS Code <a class=\"medius-aisummary\" data-starttime=\"00:00:01\">00:00:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>. He describes the plugin as a powerful integrated tool that helps developers automatically scan for code issues including secrets, security vulnerabilities, IEC compliance problems, and general code quality concerns as they write code. Simon emphasizes that this capability supports the \"shift left\" approach by catching issues early in the development process <a class=\"medius-aisummary\" data-starttime=\"00:00:19\">00:00:19</a>.</p>\n\n<p><strong>Core Features and Integrations:</strong> The plugin provides several additional functionalities beyond real-time code scanning. Simon highlights the pre-commit hook feature, which checks for exposed secrets prior to code commits <a class=\"medius-aisummary\" data-starttime=\"00:00:27\">00:00:27</a>, and the Safe Chain integration that acts as a malware package scanner <a class=\"medius-aisummary\" data-starttime=\"00:00:34\">00:00:34</a>. Another advanced component, the Aikido MCP, assists with proactive analysis such as automated checks for generated code or AI prompts to ensure they adhere to security and quality guidelines <a class=\"medius-aisummary\" data-starttime=\"00:00:39\">00:00:39</a>.</p>\n\n<p><strong>Setup and Installation Process:</strong> Simon assures viewers that installation is straightforward <a class=\"medius-aisummary\" data-starttime=\"00:00:51\">00:00:51</a>. Users simply need an Aikido license and can log in using either a login button or a personal access token <a class=\"medius-aisummary\" data-starttime=\"00:00:55\">00:00:55</a>. Once active, the plugin scans open files within VS Code and visually flags detected issues. By selecting an issue, developers can immediately locate the problematic section within the code and view Aikido’s explanation and suggested resolution <a class=\"medius-aisummary\" data-starttime=\"00:01:08\">00:01:08</a>.</p>\n\n<p><strong>AI-Powered Fixing and Workflow Integration:</strong> Aikido AI plays a central role by helping users evaluate issue impact, report false positives, or even apply AI-generated fixes automatically <a class=\"medius-aisummary\" data-starttime=\"00:01:21\">00:01:21</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:43\">00:01:43</a>. When a fix is suggested, the plugin opens a separate tab that displays a diff-style comparison — with red lines marking problematic code and green lines showing the corrected version <a class=\"medius-aisummary\" data-starttime=\"00:02:11\">00:02:11</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:16\">00:02:16</a>. Users can then choose to apply or reject the proposed changes. Simon notes that Aikido’s interactive interface makes it easy to adopt corrections and maintain safe, high-quality code directly within the IDE.</p>\n\n<p><strong>Comprehensive Scanning Capabilities:</strong> Beyond individual file analysis, Simon demonstrates how the plugin can run both dependency scans and full workspace scans <a class=\"medius-aisummary\" data-starttime=\"00:01:50\">00:01:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:30\">00:02:30</a>. The dependency scan identifies vulnerabilities in open source libraries, while the workspace scan reviews the entire repository to list all files with detected issues. Clicking on any result opens the corresponding file and reveals the flagged lines along with specific details about the problem <a class=\"medius-aisummary\" data-starttime=\"00:03:03\">00:03:03</a>. The plugin even supports automated fixes for open source dependencies, updating outdated or insecure packages directly within the IDE <a class=\"medius-aisummary\" data-starttime=\"00:03:22\">00:03:22</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:37\">00:03:37</a>.</p>\n\n<p><strong>Conclusion and Expansion:</strong> The demo concludes with Simon reiterating the versatility and ease of the Aikido VS Code plugin. With its expansion packs, users can access all of these advanced capabilities — from secret detection to automated fixes — integrated smoothly into their day-to-day development environment <a class=\"medius-aisummary\" data-starttime=\"00:03:42\">00:03:42</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:48\">00:03:48</a>. He wraps up the session by thanking viewers and wishing them a great day, leaving an impression that Aikido simplifies continuous security and quality assurance directly inside VS Code <a class=\"medius-aisummary\" data-starttime=\"00:03:49\">00:03:49</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:53\">00:03:53</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/0f705703-0b2f-42ca-871e-a21648b9f954","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/0f705703-0b2f-42ca-871e-a21648b9f954","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/0f705703-0b2f-42ca-871e-a21648b9f954","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/0f705703-0b2f-42ca-871e-a21648b9f954","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.382Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Aikido Security"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/0f705703-0b2f-42ca-871e-a21648b9f954","ogTitle":"Ship safer code from your IDE with Aikido","ogDescription":"Learn how you can integrate security scanning directly into the IDE with Aikido. Using Aikido’s extension, you can ensure that you are shipping less vulnerabilities with your PRs, are not introducing secrets into your commits, and even block malware from making it onto your machine. Best of all, it's extremely lightweight and doesn't slow you down.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Resiliency","logicalValue":"Resiliency"},{"displayValue":"Security","logicalValue":"Security"},{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Visual Studio Code","logicalValue":"Visual Studio Code"},{"displayValue":"Community","logicalValue":"Community"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"App Developers","logicalValue":"App Developers"},{"displayValue":"Platform Security","logicalValue":"Platform Security"},{"displayValue":"Secure App Development","logicalValue":"Secure App Development"},{"displayValue":"Attack Surface Reduction","logicalValue":"Attack Surface Reduction"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Agentic Security","logicalValue":"Agentic Security"},{"displayValue":"Developer Technologies","logicalValue":"Developer Technologies"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"},{"displayValue":"DevSecOps","logicalValue":"DevSecOps"}]},{"@search.score":1.0,"documentId":"620f7c75-0ce6-4d28-ad93-7b39566128d2","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780001858845001fral","langLocale":"en-US","sessionId":"1780001858845001fral","sessionInstanceId":"1780001858845001fral","sessionCode":"OD804","sessionCodeNormalized":"EN-US-OD804","title":"Simplifying .NET Installs with dotnetup","sortTitle":"simplifying .net installs with dotnetup","description":"A new way to manage .NET SDK and Runtime installations that works for every user, on every platform!","aiDescription":"<p><strong>Introduction and Problem Definition:</strong> The video opens with Chet Husk welcoming viewers to Build and introducing the session focused on simplifying .NET installations through a new tool called .NET Up <a class=\"medius-aisummary\" data-starttime=\"00:00:02\">00:00:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:08\">00:00:08</a>. He outlines the agenda—to discuss installation challenges, introduce the proposed solution, demonstrate early preview features, and share an upcoming roadmap. Chet explains that managing .NET versions and installations across different platforms is complex, especially for developers outside of Visual Studio environments <a class=\"medius-aisummary\" data-starttime=\"00:00:12\">00:00:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:06\">00:01:06</a>. He identifies the heterogeneous ecosystem of installers like package managers and manual scripts, emphasizing pain points around inconsistency, lack of management visibility, and version mismatches that plague developers when setting up .NET <a class=\"medius-aisummary\" data-starttime=\"00:04:41\">00:04:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:15\">00:07:15</a>.</p>\n\n<p><strong>Introducing .NET Up:</strong> To solve these challenges, Chet presents .NET Up—a lightweight cross-platform management tool designed for user-global installations that do not need administrative rights <a class=\"medius-aisummary\" data-starttime=\"00:07:20\">00:07:20</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:28\">00:08:28</a>. Built using native AOT for speed, .NET Up aims to unify installation and management under a consistent experience across operating systems. It leverages existing metadata like global.json to understand project SDK requirements, introduces auditability through signed binaries, and plans to integrate with CI/CD and dependency management tools like Dependabot and Renovate <a class=\"medius-aisummary\" data-starttime=\"00:09:04\">00:09:04</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:00\">00:11:00</a>. Chet transitions into internal preview demos but underscores that it’s still an early-stage project seeking community feedback <a class=\"medius-aisummary\" data-starttime=\"00:11:07\">00:11:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:11:35\">00:11:35</a>.</p>\n\n<p><strong>Demo – Basic Installation:</strong> The first demonstration shows how developers can easily acquire and initialize .NET Up using a simple script hosted via aka.ms links <a class=\"medius-aisummary\" data-starttime=\"00:12:00\">00:12:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:12\">00:13:12</a>. Once installed, the init command guides users through configuration choices such as channels and usage modes. The tool automatically downloads validated SDKs tailored for the host platform and configures profiles—setting PATH and environment variables <a class=\"medius-aisummary\" data-starttime=\"00:14:02\">00:14:02</a>–<a class=\"medius-aisummary\" data-starttime=\"00:16:33\">00:16:33</a>. Chet demonstrates running “.NET Up Info” to list installed versions, proving the installation works seamlessly. The tool’s ability to manage shell configuration and maintain user-level installations simplifies the developer experience dramatically compared to managing discrete SDKs manually <a class=\"medius-aisummary\" data-starttime=\"00:16:45\">00:16:45</a>–<a class=\"medius-aisummary\" data-starttime=\"00:17:05\">00:17:05</a>.</p>\n\n<p><strong>Demo – Project Compatibility and Multi-Target Testing:</strong> The second part showcases how .NET Up respects project-specific version constraints defined in global.json <a class=\"medius-aisummary\" data-starttime=\"00:18:08\">00:18:08</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:10\">00:19:10</a>. Chet switches between SDK versions, demonstrating automatic resolution based on that file’s specification. He then moves into an advanced multi-target test scenario, running a library compatible with .NET 8, 9, and 10 <a class=\"medius-aisummary\" data-starttime=\"00:23:15\">00:23:15</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:00\">00:27:00</a>. When certain tests fail due to missing runtimes, he installs them efficiently using “.NET Up Runtime Install” rather than multiple full SDKs <a class=\"medius-aisummary\" data-starttime=\"00:27:12\">00:27:12</a>–<a class=\"medius-aisummary\" data-starttime=\"00:29:59\">00:29:59</a>. .NET Up tracks and updates SDKs and runtimes via manifests, enabling centralized management and automated cleanup—removing versions no longer required <a class=\"medius-aisummary\" data-starttime=\"00:30:32\">00:30:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:00\">00:33:00</a>.</p>\n\n<p><strong>Feature Highlights and Future Planning:</strong> Beyond installations and updates, Chet highlights other helpful commands like “Print Init Script” and “Uninstall” <a class=\"medius-aisummary\" data-starttime=\"00:31:43\">00:31:43</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:55\">00:33:55</a>. He envisions component-level management similar to Rustup—where developers could control SDK subsets and installation cadence. The roadmap includes internal preview, public preview, and GA milestones <a class=\"medius-aisummary\" data-starttime=\"00:38:06\">00:38:06</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:02\">00:41:02</a>. Internal previews will focus on SDK management, nightly build integration, and “one-shot execution” to compare different tool versions <a class=\"medius-aisummary\" data-starttime=\"00:39:31\">00:39:31</a>–<a class=\"medius-aisummary\" data-starttime=\"00:41:00\">00:41:00</a>. Public preview will add self-updating behavior, signed validation, and integration with agent skill repositories for automation <a class=\"medius-aisummary\" data-starttime=\"00:41:19\">00:41:19</a>–<a class=\"medius-aisummary\" data-starttime=\"00:43:07\">00:43:07</a>. The GA release aims to make .NET Up official in .NET’s installation documentation and expand global.json to manage runtimes natively.</p>\n\n<p><strong>Conclusion and Call to Action:</strong> Chet concludes with next steps and resources: installation scripts, documentation, design specs, and feedback forums <a class=\"medius-aisummary\" data-starttime=\"00:46:05\">00:46:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:47:03\">00:47:03</a>. He provides repository links to demo projects and encourages viewers to clone, test, and share feedback via GitHub discussions. The team seeks user input on performance, usability, and missing features to refine .NET Up before general release. The session ends with appreciation to the Build audience and an invitation to explore, experiment, and contribute to the next evolution of .NET management tools <a class=\"medius-aisummary\" data-starttime=\"00:47:27\">00:47:27</a>–<a class=\"medius-aisummary\" data-starttime=\"00:47:46\">00:47:46</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/bb273074-b360-4724-bb2f-31167ffe2b19","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/bb273074-b360-4724-bb2f-31167ffe2b19","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/bb273074-b360-4724-bb2f-31167ffe2b19","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/bb273074-b360-4724-bb2f-31167ffe2b19","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/bb273074-b360-4724-bb2f-31167ffe2b19","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/bb273074-b360-4724-bb2f-31167ffe2b19","ogTitle":"Simplifying .NET Installs with dotnetup","ogDescription":"A new way to manage .NET SDK and Runtime installations that works for every user, on every platform!","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/dotnetup/get-started","backpackLink":null,"label":"Get started","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended Next Step","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"2bf98cff-8bfe-4899-bc5b-6dac48dd5cbf","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1776799345875001D2ow","langLocale":"en-US","sessionId":"1776799345875001D2ow","sessionInstanceId":"1776799345875001D2ow","sessionCode":"OD822","sessionCodeNormalized":"EN-US-OD822","title":"Smarter PostgreSQL migrations to power modern, intelligent apps","sortTitle":"smarter postgresql migrations to power modern, intelligent apps","description":"Building intelligent apps takes more than moving data. High-performing teams migrate early, standardize on open source, and start using their data to power real intelligence. In this session, see how teams migrate terabytes to Azure Database for PostgreSQL using AI-assisted and integrated tools that reduce risk and downtime then unlock performance, security, and AI-ready capabilities on Azure to ship smarter solutions, from AI-driven platforms to multi-tenant apps.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/be28e385-b924-43f7-a400-543b714b48ac","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/be28e385-b924-43f7-a400-543b714b48ac","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/be28e385-b924-43f7-a400-543b714b48ac","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/be28e385-b924-43f7-a400-543b714b48ac","speakerIds":[],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.384Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/be28e385-b924-43f7-a400-543b714b48ac","ogTitle":"Smarter PostgreSQL migrations to power modern, intelligent apps","ogDescription":"Building intelligent apps takes more than moving data. High-performing teams migrate early, standardize on open source, and start using their data to power real intelligence. In this session, see how teams migrate terabytes to Azure Database for PostgreSQL using AI-assisted and integrated tools that reduce risk and downtime then unlock performance, security, and AI-ready capabilities on Azure to ship smarter solutions, from AI-driven platforms to multi-tenant apps.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Azure Database for PostgreSQL","logicalValue":"Azure Database for PostgreSQL"},{"displayValue":"CP&D","logicalValue":"CP&D"}]},{"@search.score":1.0,"documentId":"f2177ba5-578e-45c9-8e81-8c6179027d6e","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Will Wang","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396264136001UEob","langLocale":"en-US","sessionId":"1774396264136001UEob","sessionInstanceId":"1774396264136001UEob","sessionCode":"ODSP908","sessionCodeNormalized":"EN-US-ODSP908","title":"Streamline your end-to-end developer workflow with AI","sortTitle":"streamline your end-to-end developer workflow with ai","description":"Most developers want to ship their own ideas, but the friction between coding, communication, and release slows them down. In this session, follow a real daily workflow building Clova—from ideation to launch—using GitHub Copilot in VS Code, Microsoft Teams, Logitech MX tools, and AI-assisted workflows. Learn practical patterns to streamline work, reduce friction, and accelerate execution.","aiDescription":"<p><strong>Introduction and Overview:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>, Will Wang, founder of Clova, welcomes viewers and introduces the topic of streamlining end-to-end developer workflows with AI. He shares background details about his startup, which helps brands generate on-brand assets efficiently, and mentions being backed by the Founders Inc. accelerator in San Francisco. Will also notes his YouTube presence with over 18,000 subscribers, where he documents his startup journey. He outlines the video’s structure, explaining that he will walk through four phases of product shipping—from ideation to building, team communication, and going to market <a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>.</p>\n\n<p><strong>Phase 1 - Generating Startup Ideas:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:00:54\">00:00:54</a>, Will advises building solutions for problems one personally experiences, emphasizing that firsthand understanding allows for rapid iteration and deeper insight. He highlights domain expertise as another asset for identifying strong startup ideas and gaining access to early adopters through industry networks. He shares that inspiration often strikes unexpectedly, recommending tools like the Logitech MX Master 4 mouse to quickly capture ideas via notepad or screenshots <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>. Will also discusses conducting unbiased user interviews focused on task-based questions—what people spend time on, their workflows, and organizational budgets—resulting in valuable, non-opinionated insights for product ideation <a class=\"medius-aisummary\" data-starttime=\"00:02:38\">00:02:38</a>.</p>\n\n<p><strong>Phase 2 - Building with AI Tools:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:02:58\">00:02:58</a>, Will transitions to the building phase, describing how he uses GitHub Copilot to write code, documentation, and tests more effectively. He demonstrates configuring pre-set prompts on his Logitech Creative MX Keypad to automate repetitive AI interactions within VS Code, preventing context switching and boosting productivity <a class=\"medius-aisummary\" data-starttime=\"00:03:26\">00:03:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:39\">00:03:39</a>. Will emphasizes managing AI sessions per feature to maintain context clarity and writes detailed AI logs to connect work across sessions. He notes that, due to the speed of modern iteration, his team now ships multiple features to production quickly—letting market feedback, rather than internal debates, determine direction <a class=\"medius-aisummary\" data-starttime=\"00:04:21\">00:04:21</a>.</p>\n\n<p><strong>Phase 3 - AI-Enhanced Team Communication:</strong> Starting around <a class=\"medius-aisummary\" data-starttime=\"00:04:36\">00:04:36</a>, Will explains how AI has dramatically improved communication within teams. Previously, he relied on verbal updates, but now he uses AI to generate structured HTML documentation that visually aligns his team during meetings. Tools like Microsoft Teams automatically capture transcripts and generate actionable summaries, enabling clearer decision-making and recordkeeping <a class=\"medius-aisummary\" data-starttime=\"00:05:00\">00:05:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:05:31\">00:05:31</a>. He maintains pre-tested prompts for consistent AI outputs, stored conveniently on his Logitech keypad for instant access. This streamlined documentation workflow saves time while improving clarity and follow-through across both technical and business discussions.</p>\n\n<p><strong>Phase 4 - Going to Market:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:05:53\">00:05:53</a>, Will turns to post-launch strategies, emphasizing finding users where they gather. For developer-focused products, he recommends Twitter and Product Hunt, while also advocating open-source projects to attract contributors and build visibility through GitHub Stars. For his own startup, Clova, Will primarily markets on Instagram and LinkedIn <a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:43\">00:06:43</a>. He shares how building in public—such as sharing his process on YouTube—creates emotional investment and organic growth. Will also describes using AI-integrated message templates for outreach, allowing him to personalize but send efficiently <a class=\"medius-aisummary\" data-starttime=\"00:07:16\">00:07:16</a>–<a class=\"medius-aisummary\" data-starttime=\"00:07:23\">00:07:23</a>.</p>\n\n<p><strong>Conclusion and Key Insights:</strong> Near the end <a class=\"medius-aisummary\" data-starttime=\"00:07:28\">00:07:28</a>, Will reflects on AI’s transformative leverage—enabling faster building, testing, and iteration of software ideas. He observes a growing trend toward personalized applications that were previously too costly to create, now made viable by AI acceleration <a class=\"medius-aisummary\" data-starttime=\"00:07:49\">00:07:49</a>. He concludes by encouraging viewers to ship their own projects and share them with him, expressing excitement about the emerging wave of creativity and innovation AI makes possible. The presentation wraps up at <a class=\"medius-aisummary\" data-starttime=\"00:08:07\">00:08:07</a> with gratitude and encouragement to continue learning and building.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/71afe298-019f-47eb-b04a-8b0630b60918","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/71afe298-019f-47eb-b04a-8b0630b60918","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/71afe298-019f-47eb-b04a-8b0630b60918","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/71afe298-019f-47eb-b04a-8b0630b60918","speakerIds":["1768245294609build26-1776428358928001Ukpm"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.64Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Logitech"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/71afe298-019f-47eb-b04a-8b0630b60918","ogTitle":"Streamline your end-to-end developer workflow with AI","ogDescription":"Most developers want to ship their own ideas, but the friction between coding, communication, and release slows them down. In this session, follow a real daily workflow building Clova—from ideation to launch—using GitHub Copilot in VS Code, Microsoft Teams, Logitech MX tools, and AI-assisted workflows. Learn practical patterns to streamline work, reduce friction, and accelerate execution.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776428358928001Ukpm","RegistrantKey":"1768245294609build26-1776428358928001Ukpm"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"GitHub Copilot","logicalValue":"GitHub Copilot"},{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"Windows Developer","logicalValue":"Windows Developer"},{"displayValue":"VS Code","logicalValue":"VS Code"},{"displayValue":"DevTools","logicalValue":"DevTools"},{"displayValue":"Dev Tools","logicalValue":"Dev Tools"}]},{"@search.score":1.0,"documentId":"6bd77cdd-9263-4596-b684-75e1561a1856","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Harry Kimpel","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774396320064001zLWO","langLocale":"en-US","sessionId":"1774396320064001zLWO","sessionInstanceId":"1774396320064001zLWO","sessionCode":"ODSP909","sessionCodeNormalized":"EN-US-ODSP909","title":"Take AI agents from prototype to production with OpenTelemetry","sortTitle":"take ai agents from prototype to production with opentelemetry","description":"Shipping an AI agent is easy. Trusting it in production is hard. In this session, build a multi-agent travel planner using the Microsoft Agent Framework, then instrument it end-to-end with OpenTelemetry and New Relic. You'll learn how to trace agent decisions, monitor response quality, catch prompt injection attacks, and build CI/CD quality gates—so your AI systems are observable, secure, and ready for real customers.","aiDescription":"<p><strong>Introduction and Problem Statement:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:00:06\">00:00:06</a>, Harry Kimpel from New Relic welcomes viewers and poses a question often asked by developers: why do AI agents behave unpredictably in production despite performing well during testing? He contrasts traditional web services, which follow clear request-response paths and allow debugging through logs, with AI agents that do not follow fixed code paths. These agents autonomously decide actions, invoke tools, and sometimes delegate tasks to sub-agents, creating complex reasoning chains. Kimpel highlights the inherent difficulty in pinpointing where issues occur within these opaque systems and stresses that without observability, teams remain blind to the cause of anomalies, leading to confusion between random flukes and systemic failures. This inability to “see inside” AI systems is identified as the core challenge the talk aims to address.</p>\n\n<p><strong>Building the Wanda AI Application:</strong> Around <a class=\"medius-aisummary\" data-starttime=\"00:02:00\">00:02:00</a>, Kimpel introduces a practical example by proposing the creation of “Wanda AI,” a travel planning startup. The idea is to build an AI-powered system that takes a user’s travel preferences and generates personalized itineraries. Investors appreciate the demo but need assurance that the agent’s recommendations are valid, performant, and debuggable. He establishes that these are fundamentally engineering issues solvable through observability. To build the system, Kimpel selects a technology stack comprising the Microsoft Agent Framework for orchestration, OpenTelemetry for instrumentation, and New Relic as the observability backend. Together, these layers will make the AI’s internal workings measurable and understandable.</p>\n\n<p><strong>Agent Framework and Instrumentation:</strong> Starting at <a class=\"medius-aisummary\" data-starttime=\"00:03:23\">00:03:23</a>, Kimpel explains the Microsoft Agent Framework, which allows the structured definition of agents and their tools within a multi-agent system. The Wanda AI setup includes a Flask web app for input, a primary travel planning agent to handle reasoning, and tool modules like destination search, weather forecast, and itinerary builder. While the framework effectively models human-like problem-solving steps at language-model speed (<a class=\"medius-aisummary\" data-starttime=\"00:05:04\">00:05:04</a>), the code remains opaque without observability. To resolve this, he integrates OpenTelemetry—the open standard for tracing, metrics, and logs. The Agent Framework’s built-in OpenTelemetry support simplifies instrumentation, enabling automatic emission of traces with timestamps and status codes. Once enabled (<a class=\"medius-aisummary\" data-starttime=\"00:06:36\">00:06:36</a>), every agent invocation becomes a trace that visualizes the reasoning flow in tools like New Relic.</p>\n\n<p><strong>Visualizing and Extending Observability:</strong> At <a class=\"medius-aisummary\" data-starttime=\"00:07:01\">00:07:01</a>, Kimpel reviews an actual Wanda AI trace in New Relic. It shows 48.2 seconds total duration with clear breakdowns—destination selection taking 322 ms, weather forecast 1.17 s, and itinerary builder nearly 39 s—immediately exposing performance bottlenecks. Built-in telemetry reveals “what” happened but not “why,” prompting the addition of custom spans and metrics for enriched context (<a class=\"medius-aisummary\" data-starttime=\"00:07:55\">00:07:55</a>). These custom spans might track business-relevant data like destination categories, enabling advanced filtering and insight into user preferences. Custom metrics like itinerary counts and cache hit rates provide statistical visibility. He continues with logging best practices, explaining that simple print statements are insufficient. With OpenTelemetry, logs can include trace and span IDs, allowing direct correlation between log entries and traces (<a class=\"medius-aisummary\" data-starttime=\"00:09:10\">00:09:10</a>). In New Relic, this linkage enables developers to click from an error log straight to the corresponding trace, identifying root causes in under a minute instead of manually matching entries across different tools.</p>\n\n<p><strong>Quality Gates and Security Controls:</strong> Beginning at <a class=\"medius-aisummary\" data-starttime=\"00:10:12\">00:10:12</a>, Kimpel emphasizes production-readiness through quality assurance and security. He introduces evaluation tests that act like AI behavior unit tests, scoring outputs against expected results for realism, relevancy, and format correctness. Wired into CI/CD pipelines (<a class=\"medius-aisummary\" data-starttime=\"00:11:02\">00:11:02</a>), these evaluations automatically block deployments if quality drops below thresholds. For security, particularly prompt injection defenses (<a class=\"medius-aisummary\" data-starttime=\"00:11:28\">00:11:28</a>), he details two protection layers. Microsoft Foundry guardrails operate at the platform level, catching explicit and indirect attacks. Application-level detection examines incoming requests for malicious patterns. Both layers are instrumented via OpenTelemetry, making injection attempts, patterns, and guardrail activations observable in New Relic dashboards. This turns security—a traditionally opaque aspect—into data that can be monitored and analyzed alongside performance metrics.</p>\n\n<p><strong>Conclusion and Next Steps:</strong> In the closing section (<a class=\"medius-aisummary\" data-starttime=\"00:13:29\">00:13:29</a>), Kimpel revisits the initial query about unpredictable AI behavior. The solution, he reiterates, lies in observability: enabling visibility into agents through integrated tracing, logging, and quality controls. He summarizes the build of Wanda AI using Microsoft Agent Framework, OpenTelemetry, and New Relic, emphasizing that the same approach can be applied universally to agent-based systems. Key takeaways include starting telemetry early, adding custom instrumentation for business context, correlating logs, enforcing quality gates, and monitoring security layers (<a class=\"medius-aisummary\" data-starttime=\"00:14:51\">00:14:51</a>). The session closes by pointing viewers to the Microsoft “What the Hack” repository (<a class=\"medius-aisummary\" data-starttime=\"00:15:01\">00:15:01</a>), specifically hack number 073 focused on New Relic agent observability, providing hands-on exercises and full source code for reproducing the presented scenario. Kimpel ends by encouraging teams to “ship observable AI,” reinforcing observability as the foundation for reliable and secure AI deployment.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/05becd32-00b0-410e-adb0-4dc892dbde19","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/05becd32-00b0-410e-adb0-4dc892dbde19","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/05becd32-00b0-410e-adb0-4dc892dbde19","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/05becd32-00b0-410e-adb0-4dc892dbde19","speakerIds":["1768245294609build26-1778029118927001DnBr"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.64Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["New Relic"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/05becd32-00b0-410e-adb0-4dc892dbde19","ogTitle":"Take AI agents from prototype to production with OpenTelemetry","ogDescription":"Shipping an AI agent is easy. Trusting it in production is hard. In this session, build a multi-agent travel planner using the Microsoft Agent Framework, then instrument it end-to-end with OpenTelemetry and New Relic. You'll learn how to trace agent decisions, monitor response quality, catch prompt injection attacks, and build CI/CD quality gates—so your AI systems are observable, secure, and ready for real customers.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Agents & apps","logicalValue":"Agents & apps"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778029118927001DnBr","RegistrantKey":"1768245294609build26-1778029118927001DnBr"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Observability","logicalValue":"Observability"},{"displayValue":"Reliability","logicalValue":"Reliability"},{"displayValue":"Monitor","logicalValue":"Monitor"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Azure DevOps","logicalValue":"Azure DevOps"},{"displayValue":"Agent Observability","logicalValue":"Agent Observability"},{"displayValue":"Developer Frameworks","logicalValue":"Developer Frameworks"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"}]},{"@search.score":1.0,"documentId":"1f1cf2e9-a073-41f9-81db-06060f6fcd95","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Gerald Versluis","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1780001184901001vDCR","langLocale":"en-US","sessionId":"1780001184901001vDCR","sessionInstanceId":"1780001184901001vDCR","sessionCode":"OD803","sessionCodeNormalized":"EN-US-OD803","title":"Taking your AI to the edge with .NET MAUI","sortTitle":"taking your ai to the edge with .net maui","description":"AI is transforming both what we build and how we build it. Learn how .NET MAUI developers can bring AI to the edge using local models and on-device capabilities across mobile and desktop, while understanding the role of cloud AI. We’ll cover the impact on privacy, performance, and UX, explore .NET 10 features and what’s coming in .NET 11, and show how AI-powered tools and agentic workflows like DevFlow can accelerate app development.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/bfd5191f-c11a-4198-864b-8f36010da199","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/bfd5191f-c11a-4198-864b-8f36010da199","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/bfd5191f-c11a-4198-864b-8f36010da199","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/bfd5191f-c11a-4198-864b-8f36010da199","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/bfd5191f-c11a-4198-864b-8f36010da199","speakerIds":["1768245294609build26-1780001467107001fsR5"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:37:01.46Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/bfd5191f-c11a-4198-864b-8f36010da199","ogTitle":"Taking your AI to the edge with .NET MAUI","ogDescription":"AI is transforming both what we build and how we build it. Learn how .NET MAUI developers can bring AI to the edge using local models and on-device capabilities across mobile and desktop, while understanding the role of cloud AI. We’ll cover the impact on privacy, performance, and UX, explore .NET 10 features and what’s coming in .NET 11, and show how AI-powered tools and agentic workflows like DevFlow can accelerate app development.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[{"displayValue":"C#","logicalValue":"C#"}],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1780001467107001fsR5","RegistrantKey":"1768245294609build26-1780001467107001fsR5"}],"repeatedSessions":[],"relatedResources":[{"link":"https://aka.ms/maui","backpackLink":null,"label":"Learn more","category":"Web link"}],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended Next Step","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":".NET","logicalValue":".NET"},{"displayValue":"Developer","logicalValue":"Developer"}]},{"@search.score":1.0,"documentId":"90ab61db-1f51-4ac8-925c-c7e4bc90e0d7","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Piero Morano, Raki Rahman","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693229404001XedV","langLocale":"en-US","sessionId":"1775693229404001XedV","sessionInstanceId":"1775693229404001XedV","sessionCode":"OD818","sessionCodeNormalized":"EN-US-OD818","title":"The AI-native Data Engineer","sortTitle":"the ai-native data engineer","description":"This session explores how AI is transforming data engineering. Discover how agentic Copilot across Fabric Web, VS Code, and CLI removes coding barriers, automates repetitive tasks, and lets engineers focus on optimizing data stacks to unlock faster insights. We’ll also cover the latest Fabric releases and how they accelerate the path from idea to production.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"https://medius.microsoft.com/video/asset/PPT/6abe78db-035f-433c-a14d-78dc6336809e","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/6abe78db-035f-433c-a14d-78dc6336809e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/6abe78db-035f-433c-a14d-78dc6336809e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/6abe78db-035f-433c-a14d-78dc6336809e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/6abe78db-035f-433c-a14d-78dc6336809e","speakerIds":["1768245294609build26-1777915932896001N9El","1768245294609build26-1777915963063001tX3H"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.896Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/6abe78db-035f-433c-a14d-78dc6336809e","ogTitle":"The AI-native Data Engineer","ogDescription":"This session explores how AI is transforming data engineering. Discover how agentic Copilot across Fabric Web, VS Code, and CLI removes coding barriers, automates repetitive tasks, and lets engineers focus on optimizing data stacks to unlock faster insights. We’ll also cover the latest Fabric releases and how they accelerate the path from idea to production.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777915932896001N9El","RegistrantKey":"1768245294609build26-1777915932896001N9El"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1777915963063001tX3H","RegistrantKey":"1768245294609build26-1777915963063001tX3H"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"9b218b5d-7f7c-4495-8f56-f4b6b26b4fd5","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Dipti Borkar, Joshua Caplan, Miquella de Boer, Wee hyong Tok","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1775693160083001xdWH","langLocale":"en-US","sessionId":"1775693160083001xdWH","sessionInstanceId":"1775693160083001xdWH","sessionCode":"OD815","sessionCodeNormalized":"EN-US-OD815","title":"Unify your entire data estate on a single, AI-ready data lake","sortTitle":"unify your entire data estate on a single, ai-ready data lake","description":"Every AI initiative starts with the same fundamental challenge: understanding where your data lives and how to bring it together. Microsoft OneLake solves this by unifying data across clouds, on-premises environments, and third-party platforms into a single logical data lake without unnecessary ETL, fragmentation, or duplication. Join this session to explore new OneLake sources, see how OneLake now supports data projects outside of Fabric, and how OneLake integrates with Foundry for AI projects.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/7ac6338c-2c31-4bce-b01d-198ee858c97e","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/7ac6338c-2c31-4bce-b01d-198ee858c97e","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/7ac6338c-2c31-4bce-b01d-198ee858c97e","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/7ac6338c-2c31-4bce-b01d-198ee858c97e","speakerIds":["1768245294609build26-1776284363899001XcpL","1768245294609build26-1776284546293001iQot","1768245294609build26-1776284496474001NOwD","1768245294609build26-1776284522426001x22N"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.895Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/7ac6338c-2c31-4bce-b01d-198ee858c97e","ogTitle":"Unify your entire data estate on a single, AI-ready data lake","ogDescription":"Every AI initiative starts with the same fundamental challenge: understanding where your data lives and how to bring it together. Microsoft OneLake solves this by unifying data across clouds, on-premises environments, and third-party platforms into a single logical data lake without unnecessary ETL, fragmentation, or duplication. Join this session to explore new OneLake sources, see how OneLake now supports data projects outside of Fabric, and how OneLake integrates with Foundry for AI projects.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Cloud platform & data","logicalValue":"Cloud platform & data"}],"sessionLevel":[{"displayValue":"(200) Intermediate","logicalValue":"(200) Intermediate"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284496474001NOwD","RegistrantKey":"1768245294609build26-1776284496474001NOwD"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284522426001x22N","RegistrantKey":"1768245294609build26-1776284522426001x22N"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284546293001iQot","RegistrantKey":"1768245294609build26-1776284546293001iQot"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1776284363899001XcpL","RegistrantKey":"1768245294609build26-1776284363899001XcpL"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Microsoft Fabric","logicalValue":"Microsoft Fabric"},{"displayValue":"CP&D","logicalValue":"CP&D"},{"displayValue":"Data","logicalValue":"Data"}]},{"@search.score":1.0,"documentId":"025df61d-5db5-40c9-9a34-601a97a51d54","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Keiji Kanazawa, Caroline Matthews","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1777073860295001Venk","langLocale":"en-US","sessionId":"1777073860295001Venk","sessionInstanceId":"1777073860295001Venk","sessionCode":"ODSP934","sessionCodeNormalized":"EN-US-ODSP934","title":"Unlock Claude in Microsoft Foundry","sortTitle":"unlock claude in microsoft foundry","description":"This session introduces enterprise developers and IT leaders to Claude models now available in Microsoft Foundry. Hear from Anthropic and Microsoft experts on deploying frontier AI within existing Azure infrastructure to build agents and automate complex workflows. See a live demo of Claude in action, plus a walkthrough of Claude model selection and how to get started with billing through MACC.","aiDescription":"<p><strong>Introduction and Business Context:</strong> The webinar opens with Keiji Kanazawa welcoming viewers to the Microsoft and Anthropic session on building intelligent agents using Claude within Microsoft Foundry (<a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:13\">00:00:13</a>). He introduces the idea that leveraging AI to meet business needs requires understanding how intelligent systems can foster growth, enhance customer satisfaction, and optimize workflows. Caroline Matthews from Anthropic joins, describing her role in supporting teams deploying Claude in production (<a class=\"medius-aisummary\" data-starttime=\"00:00:24\">00:00:24</a>–<a class=\"medius-aisummary\" data-starttime=\"00:00:35\">00:00:35</a>). Kanazawa outlines the foundational needs for enterprise AI: orchestrating multiple agents, ensuring interoperability, integrating with data sources like Microsoft 365, and maintaining continuous monitoring with robust safety and evaluation tools (<a class=\"medius-aisummary\" data-starttime=\"00:00:51\">00:00:51</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:54\">00:01:54</a>). He stresses the importance of agents that can learn, customize, and evolve while remaining secure and enterprise-ready.</p>\n\n<p><strong>Understanding Agentic Applications:</strong> The discussion then shifts to explain the nature of “agentic” applications—systems that not only generate responses but engage in long-running, goal-driven reasoning loops (<a class=\"medius-aisummary\" data-starttime=\"00:02:26\">00:02:26</a>–<a class=\"medius-aisummary\" data-starttime=\"00:03:57\">00:03:57</a>). Kanazawa illustrates how modern agents automate tasks like analyzing performance spikes and generating reports, which previously required manual effort. These agents coordinate across multiple tools, handle complex data sources, and employ memory and self-improvement cycles to refine their performance (<a class=\"medius-aisummary\" data-starttime=\"00:03:32\">00:03:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:04:46\">00:04:46</a>). He emphasizes that building, deploying, and operating these agents requires secure governance, continuous optimization, and an iterative lifecycle approach. The rationale behind the Microsoft–Anthropic partnership becomes clear: to blend Microsoft’s enterprise infrastructure with Anthropic’s deep expertise in agentic AI, harnessing leading models like Claude Opus for real-world deployment (<a class=\"medius-aisummary\" data-starttime=\"00:05:41\">00:05:41</a>–<a class=\"medius-aisummary\" data-starttime=\"00:06:50\">00:06:50</a>).</p>\n\n<p><strong>Microsoft Foundry and Claude Integration:</strong> Kanazawa introduces Microsoft Foundry as the unified control plane for creating, managing, and securing AI agents and workflows (<a class=\"medius-aisummary\" data-starttime=\"00:06:54\">00:06:54</a>–<a class=\"medius-aisummary\" data-starttime=\"00:08:17\">00:08:17</a>). Foundry integrates coding tools like GitHub Copilot and VS Code, while supporting models such as Claude Opus, Sonnet, and Haiku. Agents are governed with Microsoft Entra, Purview, and Defender, ensuring compliance and security. Caroline Matthews then elaborates on the advantages of running Claude within Microsoft’s environment (<a class=\"medius-aisummary\" data-starttime=\"00:08:32\">00:08:32</a>–<a class=\"medius-aisummary\" data-starttime=\"00:10:00\">00:10:00</a>). She outlines four pillars: frontier intelligence for challenging reasoning and coding tasks, built-in safety through Anthropic’s constitutional AI, native integration across Microsoft’s ecosystem, and a streamlined path from prototype to production. This enables teams to focus on delivering value rather than building the surrounding AI infrastructure.</p>\n\n<p><strong>The Claude Model Family and Demo:</strong> Matthews presents the Claude model lineup available in Foundry—Opus for deep reasoning, Sonnet as the robust default for enterprise work, and Haiku for high-speed, low-latency processing (<a class=\"medius-aisummary\" data-starttime=\"00:11:13\">00:11:13</a>–<a class=\"medius-aisummary\" data-starttime=\"00:13:59\">00:13:59</a>). To illustrate these concepts, she walks through an interactive “cupcake factory” demo. Using Foundry’s playground, she deploys Claude Sonnet 4.6 and personalizes it as a chatbot that accepts cupcake orders through humorous prompts (<a class=\"medius-aisummary\" data-starttime=\"00:14:10\">00:14:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:15:49\">00:15:49</a>). She then transitions to VS Code to demonstrate the Anthropic Foundry client and an MCP (Model Context Protocol) connection enabling real-time order handling. Through a live exchange, the demo shows the agent setting up a customer, retrieving information, and confirming an order—illustrating how agents can seamlessly connect external and internal systems through Foundry’s SDK tools (<a class=\"medius-aisummary\" data-starttime=\"00:15:50\">00:15:50</a>–<a class=\"medius-aisummary\" data-starttime=\"00:19:32\">00:19:32</a>).</p>\n\n<p><strong>Advanced Agent Architecture and SDK Insights:</strong> Kanazawa returns to define agent harnesses—the framework that drives how frontier models like Claude iterate, reason, and act in a continuous loop (<a class=\"medius-aisummary\" data-starttime=\"00:20:01\">00:20:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:21:06\">00:21:06</a>). Foundry’s Agent Service is highlighted as an open, secure, and multi-agent environment that supports diverse frameworks and deployment targets including M365 and Agent 365. A live demo follows, showing a customer-support agent executing automated weekly operations and generating analytical reports through the Claude Agent SDK (<a class=\"medius-aisummary\" data-starttime=\"00:22:00\">00:22:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:24:39\">00:24:39</a>). Matthews then unpacks the Claude Agent SDK itself—a production-proven harness programmable in Python and TypeScript. She explains its core loop of context gathering, action, verification, and repetition, along with built-in execution tools, permissions, lifecycle hooks, and observability features (<a class=\"medius-aisummary\" data-starttime=\"00:25:05\">00:25:05</a>–<a class=\"medius-aisummary\" data-starttime=\"00:27:04\">00:27:04</a>). The SDK lets organizations skip infrastructure plumbing and focus on building differentiated value-driven agents.</p>\n\n<p><strong>Evaluation, Observability, and Conclusion:</strong> The final part delves into observability, monitoring, and systematic evaluations to ensure quality, safety, and reliability of enterprise AI agents (<a class=\"medius-aisummary\" data-starttime=\"00:32:01\">00:32:01</a>–<a class=\"medius-aisummary\" data-starttime=\"00:33:49\">00:33:49</a>). Matthews emphasizes setting up evaluations early, covering aspects like groundedness, coherence, relevance, risk mitigation, and code robustness. These metrics, built into Azure Foundry, allow continuous monitoring across production environments, ensuring adherence to standards and identifying drift or degradation over time (<a class=\"medius-aisummary\" data-starttime=\"00:34:10\">00:34:10</a>–<a class=\"medius-aisummary\" data-starttime=\"00:35:12\">00:35:12</a>). Closing remarks feature customer outcomes from partners like Adobe and Dentons, underlining that high reasoning quality is translating into tangible business impact. Kanazawa concludes by reinforcing that combining the advanced reasoning of Claude with Microsoft Foundry’s enterprise architecture empowers teams to design resilient, scalable, and transformative AI workflows (<a class=\"medius-aisummary\" data-starttime=\"00:36:07\">00:36:07</a>–<a class=\"medius-aisummary\" data-starttime=\"00:36:36\">00:36:36</a>).</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/247dbe41-92a0-46c1-9426-8834c4c349d5","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/247dbe41-92a0-46c1-9426-8834c4c349d5","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/247dbe41-92a0-46c1-9426-8834c4c349d5","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/247dbe41-92a0-46c1-9426-8834c4c349d5","speakerIds":["1768245294609build26-1766175567403001jGYO","1768245294609build26-1778020065215001t4NN"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:56.388Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":["Anthropic","PBC"],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/247dbe41-92a0-46c1-9426-8834c4c349d5","ogTitle":"Unlock Claude in Microsoft Foundry","ogDescription":"This session introduces enterprise developers and IT leaders to Claude models now available in Microsoft Foundry. Hear from Anthropic and Microsoft experts on deploying frontier AI within existing Azure infrastructure to build agents and automate complex workflows. See a live demo of Claude in action, plus a walkthrough of Claude model selection and how to get started with billing through MACC.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Working with models","logicalValue":"Working with models"}],"sessionLevel":[{"displayValue":"(100) Foundational","logicalValue":"(100) Foundational"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778020065215001t4NN","RegistrantKey":"1768245294609build26-1778020065215001t4NN"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1766175567403001jGYO","RegistrantKey":"1768245294609build26-1766175567403001jGYO"}],"repeatedSessions":[],"relatedResources":[],"nextStep":null,"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"AI","logicalValue":"AI"},{"displayValue":"Agents","logicalValue":"Agents"},{"displayValue":"Developer","logicalValue":"Developer"},{"displayValue":"Microsoft Foundry","logicalValue":"Microsoft Foundry"},{"displayValue":"Responsible AI","logicalValue":"Responsible AI"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Foundry Agents","logicalValue":"Foundry Agents"},{"displayValue":"Agentic SDLC","logicalValue":"Agentic SDLC"},{"displayValue":"Enterprise","logicalValue":"Enterprise"}]},{"@search.score":1.0,"documentId":"ad5b57cf-5b6e-4270-8b10-faf49d06ce30","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Deepthi Chelupati, Vyom Nagrani","canUserModifySchedule":false,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1778013041569001KuG7","langLocale":"en-US","sessionId":"1778013041569001KuG7","sessionInstanceId":"1778013041569001KuG7","sessionCode":"OD800","sessionCodeNormalized":"EN-US-OD800","title":"Using autonomous SRE to move from alerts to action","sortTitle":"using autonomous sre to move from alerts to action","description":"What happens when AI agents don’t just observe incidents—but act on them? Discover how autonomous remediation, SLO management, and deep observability are redefining modern SRE workflows and changing how teams operate large scale production systems.","aiDescription":"<p><strong>Introduction and Context Setting:</strong> The session opens with background music <a class=\"medius-aisummary\" data-starttime=\"00:00:00\">00:00:00</a>–<a class=\"medius-aisummary\" data-starttime=\"00:01:29\">00:01:29</a> before hosts Vyom Nagrani and Deepthi Chelupati introduce themselves as Product Managers from the Azure SRE Agent team <a class=\"medius-aisummary\" data-starttime=\"00:01:30\">00:01:30</a>. They pose a central question: what if AI agents in site reliability engineering could not only observe incidents but actively resolve them <a class=\"medius-aisummary\" data-starttime=\"00:01:43\">00:01:43</a>? The discussion highlights the growing role of generative AI in all aspects of work, particularly in software development, where by 2028 nearly 90% of enterprise engineers are projected to use AI assistance <a class=\"medius-aisummary\" data-starttime=\"00:02:19\">00:02:19</a>. Vyom contextualizes the evolution of developer responsibilities, emphasizing how modern developers rely on agents for migration, security, release management, and documentation, while citing tangible results from companies using AI agents to improve efficiency and reduce noise in operations <a class=\"medius-aisummary\" data-starttime=\"00:02:25\">00:02:25</a>. This sets the stage for introducing Azure SRE Agent as an AI-powered system focused on operational automation, root cause diagnosis, and customization for production reliability <a class=\"medius-aisummary\" data-starttime=\"00:04:01\">00:04:01</a>.</p>\n\n<p><strong>First Demo – Preventing Issues and Reducing Toil:</strong> As the first major scenario begins <a class=\"medius-aisummary\" data-starttime=\"00:06:09\">00:06:09</a>, Deepthi demonstrates how developers can use SRE Agent to automate deployment checks and prevent regressions before they reach production. She describes typical manual tasks like code reviews and staging tests <a class=\"medius-aisummary\" data-starttime=\"00:06:32\">00:06:32</a>, then shows how the agent connects to telemetry systems such as Dynatrace, Log Analytics, and Azure activity logs <a class=\"medius-aisummary\" data-starttime=\"00:07:52\">00:07:52</a>. Through a CLI-based walkthrough, she creates an agent using a prebuilt Microsoft recipe <a class=\"medius-aisummary\" data-starttime=\"00:10:06\">00:10:06</a>, configures connectors, and sets an HTTP trigger for GitHub Actions. The trigger enables the agent to analyze every new pull request automatically, performing canary tests to identify risky changes. Deepthi validates the configuration via Copilot CLI <a class=\"medius-aisummary\" data-starttime=\"00:16:00\">00:16:00</a> and runs a demo PR to show the agent identifying a problematic database URL modification that could disrupt production <a class=\"medius-aisummary\" data-starttime=\"00:22:03\">00:22:03</a>. The agent then comments directly on the PR and flags it as high risk, streamlining error prevention and saving engineering time.</p>\n\n<p><strong>Core Capabilities and Extensibility:</strong> Vyom resumes <a class=\"medius-aisummary\" data-starttime=\"00:24:18\">00:24:18</a> by summarizing three pillars of the SRE Agent’s design—root cause analysis, automated mitigation, and incident response. The agent conducts data-driven analyses across metrics and logs, arriving at transparent, explainable insights in minutes <a class=\"medius-aisummary\" data-starttime=\"00:24:30\">00:24:30</a>. It can autonomously execute safe, reversible mitigation actions <a class=\"medius-aisummary\" data-starttime=\"00:25:01\">00:25:01</a> while keeping teams in control. Additionally, the system enhances incident response by synthesizing alerts and likely causes instantly when production issues arise <a class=\"medius-aisummary\" data-starttime=\"00:25:21\">00:25:21</a>. Vyom highlights its extensibility—organizations can embed custom logic, troubleshooting workflows, or proprietary compliance rules so that the agent behaves according to internal standards <a class=\"medius-aisummary\" data-starttime=\"00:26:00\">00:26:00</a>. Deep integration with GitHub Copilot creates a feedback loop between issue detection and development, ensuring continuous operational improvement and aligned automation.</p>\n\n<p><strong>Second Demo – Automated Incident Response:</strong> Deepthi transitions into the second scenario, demonstrating how SRE Agent handles live production incidents autonomously <a class=\"medius-aisummary\" data-starttime=\"00:27:13\">00:27:13</a>. She simulates an error by rotating a database password that was not updated in the deployed application <a class=\"medius-aisummary\" data-starttime=\"00:31:05\">00:31:05</a>. Upon receiving a ServiceNow incident alert, the agent initiates investigation, referencing stored app architecture and past incident memory to narrow down probable causes <a class=\"medius-aisummary\" data-starttime=\"00:32:00\">00:32:00</a>. It correlates logs, telemetry, and detected password rotation events, confirming them as the root cause <a class=\"medius-aisummary\" data-starttime=\"00:33:19\">00:33:19</a>. With autonomous remediation enabled, the agent updates credentials, redeploys affected containers, validates the fix, and creates a detailed GitHub issue linking all analysis for traceability <a class=\"medius-aisummary\" data-starttime=\"00:37:02\">00:37:02</a>. This example illustrates how the system can shrink mean-time-to-resolution from hours to minutes while offering configurable human oversight if preferred.</p>\n\n<p><strong>Trust, Control, and Operational Modes:</strong> Continuing from the demos <a class=\"medius-aisummary\" data-starttime=\"00:37:30\">00:37:30</a>, Vyom outlines the safeguards and enterprise controls built into Azure SRE Agent. It operates under managed identities with three permission tiers—reader, user, and administrator—to ensure least privilege access <a class=\"medius-aisummary\" data-starttime=\"00:37:45\">00:37:45</a>. Governance is enforced through configurable run modes such as review or autonomous execution <a class=\"medius-aisummary\" data-starttime=\"00:38:32\">00:38:32</a>, validated commands, and audited sessions within Application Insights <a class=\"medius-aisummary\" data-starttime=\"00:39:06\">00:39:06</a>. Security architecture eliminates secrets and uses micro-VM sandboxes, OAuth connectors, and egress proxy for controlled network access <a class=\"medius-aisummary\" data-starttime=\"00:39:15\">00:39:15</a>. Monitoring tools provide metrics dashboards and performance scores, helping teams evaluate mitigation quality and agent effectiveness <a class=\"medius-aisummary\" data-starttime=\"00:40:05\">00:40:05</a>. Vyom then categorizes three operational modes—interactive (manual chat), reactive (alert-driven response), and proactive (scheduled autonomous checks)—covering troubleshooting to automated maintenance <a class=\"medius-aisummary\" data-starttime=\"00:40:50\">00:40:50</a>.</p>\n\n<p><strong>Conclusion and Getting Started:</strong> Wrapping up, Vyom summarizes why Azure SRE Agent stands out <a class=\"medius-aisummary\" data-starttime=\"00:42:16\">00:42:16</a>: it is Azure-native, adaptable to any environment, secure by design, and integrates seamlessly with tools like GitHub, ServiceNow, PagerDuty, and Datadog. He encourages teams of all sizes to implement the system, explaining three onboarding steps—teach it with organizational runbooks <a class=\"medius-aisummary\" data-starttime=\"00:43:05\">00:43:05</a>, connect it with telemetry sources, and let it work autonomously <a class=\"medius-aisummary\" data-starttime=\"00:43:34\">00:43:34</a>. The session concludes with a call to action: visit sre.azure.com for documentation, labs, and support <a class=\"medius-aisummary\" data-starttime=\"00:43:51\">00:43:51</a>. Music plays as the video ends, reinforcing the theme of continuous AI-driven reliability and innovation in Azure operations <a class=\"medius-aisummary\" data-starttime=\"00:44:15\">00:44:15</a>.</p>","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":true,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/dc74755a-a98c-4763-b953-d49a257b8f97","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/dc74755a-a98c-4763-b953-d49a257b8f97","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/dc74755a-a98c-4763-b953-d49a257b8f97","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/dc74755a-a98c-4763-b953-d49a257b8f97","speakerIds":["1768245294609build26-1778642541081001DHZz","1768245294609build26-1765985795043001vXKc"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:57.768Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/dc74755a-a98c-4763-b953-d49a257b8f97","ogTitle":"Using autonomous SRE to move from alerts to action","ogDescription":"What happens when AI agents don’t just observe incidents—but act on them? Discover how autonomous remediation, SLO management, and deep observability are redefining modern SRE workflows and changing how teams operate large scale production systems.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Developer tools & frameworks","logicalValue":"Developer tools & frameworks"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1765985795043001vXKc","RegistrantKey":"1768245294609build26-1765985795043001vXKc"},{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1778642541081001DHZz","RegistrantKey":"1768245294609build26-1778642541081001DHZz"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[]},{"@search.score":1.0,"documentId":"5fc7683c-35b3-4bde-84c2-bc55dddfb770","startDateTime":null,"partnerType":[],"industry":[],"businessFocus":[],"product":[],"lineOfBusiness":"","speakerNames":"Samantha Song","canUserModifySchedule":true,"hasFailoverVideoUrl":false,"ssr":"","excludeFromSessionScheduler":false,"doNotPublishVideo":false,"localizedId":"en-US-1774550925443001nKS5","langLocale":"en-US","sessionId":"1774550925443001nKS5","sessionInstanceId":"1774550925443001nKS5","sessionCode":"OD858","sessionCodeNormalized":"EN-US-OD858","title":"Your PC, Your Rules: AI Agents Meet Windows Personalization","sortTitle":"your pc, your rules: ai agents meet windows personalization","description":"What if Windows personalization was as simple as asking for a starry night vibe for late night coding and the system responded with custom RGB settings? This session shows how natural language prompts map to real Windows personalization APIs. You learn how to use MCP servers that drive Dynamic Lighting, system sounds, cursor themes, and multi monitor layouts, with a clear flow from prompt to MCP schema to API call, including scope, permissions, and rollback.","aiDescription":"","businessOutcome":"","businessOutcomeTwo":"","registrationLink":"","targetAudience":"","contentArea":[],"sortRank":9,"TimeSlot":"","startDateTimeNullGroup":0,"endDateTime":null,"durationInMinutes":0,"slideDeck":"","isMandatory":false,"hasLiveStream":false,"onDemand":"https://medius.microsoft.com/Embed/video-nc/4194f3f9-a19e-4a01-a64f-d1cebedd6fd0","hasOnDemand":true,"downloadVideoLink":"https://medius.microsoft.com/video/asset/HIGHMP4/4194f3f9-a19e-4a01-a64f-d1cebedd6fd0","captionFileLink":"https://medius.microsoft.com/video/asset/Transcript/4194f3f9-a19e-4a01-a64f-d1cebedd6fd0","onDemandThumbnail":"https://medius.microsoft.com/video/asset/Thumbnail/4194f3f9-a19e-4a01-a64f-d1cebedd6fd0","speakerIds":["1768245294609build26-1774285002942001pAZM"],"sessionLinks":[],"relatedSessionCodes":[],"hasTeamsMeeting":false,"hasTeamsLiveEvent":false,"lastUpdate":"2026-07-21T19:36:52.843Z","heroSession":false,"restrictedScopes":[],"associatedCompanies":[],"captionLanguage":[],"audioLanguage":[],"aslSupported":false,"isExternalUrlSession":false,"isPopular":false,"ogImage":"https://medius.microsoft.com/video/asset/Thumbnail/4194f3f9-a19e-4a01-a64f-d1cebedd6fd0","ogTitle":"Your PC, Your Rules: AI Agents Meet Windows Personalization","ogDescription":"What if Windows personalization was as simple as asking for a starry night vibe for late night coding and the system responded with custom RGB settings? This session shows how natural language prompts map to real Windows personalization APIs. You learn how to use MCP servers that drive Dynamic Lighting, system sounds, cursor themes, and multi monitor layouts, with a clear flow from prompt to MCP schema to API call, including scope, permissions, and rollback.","deliveryTypes":[{"displayValue":"Online","logicalValue":"Online"}],"sessionType":{"displayValue":"Pre-recorded","logicalValue":"Pre-recorded"},"topic":[{"displayValue":"Windows","logicalValue":"Windows"}],"sessionLevel":[{"displayValue":"(300) Advanced","logicalValue":"(300) Advanced"}],"programmingLanguages":[],"location":{"displayValue":"","logicalValue":""},"theme":[],"solutionArea":[],"communityTopic":[],"learningZoneTopic":[],"SpeakerTypes":[{"SpeakerType":"Speaker Digital","SpeakerId":"1768245294609build26-1774285002942001pAZM","RegistrantKey":"1768245294609build26-1774285002942001pAZM"}],"repeatedSessions":[],"relatedResources":[],"nextStep":{"link":"https://aka.ms/build26-next-steps","backpackLink":null,"label":"Recommended next steps","category":"Code sample"},"event":[],"sessionLanguage":[],"viewingOptions":[{"displayValue":"Will be recorded","logicalValue":"Will be recorded"}],"audienceTypes":[],"tags":[{"displayValue":"Windows","logicalValue":"Windows"},{"displayValue":"MCP","logicalValue":"MCP"},{"displayValue":"Agents on Windows","logicalValue":"Agents on Windows"},{"displayValue":"Personalization","logicalValue":"Personalization"},{"displayValue":"Windows APIs","logicalValue":"Windows APIs"},{"displayValue":"Dynamic Lighting","logicalValue":"Dynamic Lighting"}]}]