Run and Scale AI Applications with Azure Kubernetes Service (AKS) @MicrosoftReactor
Run and Scale AI Applications with Azure Kubernetes Service (AKS)  @MicrosoftReactor
Uploaded September 2026 | Updated November 2026, 2 hours ago
Building AI applications means running demanding workloads on the right hardware, at scale, without adding operational drag.

This session covers what's new in AKS for teams building and running AI apps and agents, and how the latest capabilities help you get more out of your infrastructure while keeping day-to-day operations manageable.

You'll learn how to: - Run AI and data-intensive workloads across a wider range of accelerators, with better visibility into how they perform and where you can tune them - Scale multi-service and agentic applications across clusters and fleets, with more efficient access to the compute you need - Automate upgrades, rollouts, and configuration across your estate with more control and less risk - Connect AKS to the broader Azure AI stack to move from model to production faster, with identity and security built in Whether you run a single cluster or a large fleet, you'll leave with a clear picture of where AKS is headed for AI workloads and how to put these new capabilities to work.

đź”— https://aka.ms/AzureContainerAppsSandbox

📌 This event is a part of a series, learn more here: https://aka.ms/BCNA


0:00 Introduction and Welcome
1:13 Why run AI workloads on Kubernetes?
5:40 Key types of AI workloads: Training, Inference, and Agents
14:50 Deploying and upgrading AI apps with confidence
17:53 Demo: Developing and deploying an agentic app on AKS
26:48 Strategies for safe platform upgrades (Blue-Green/Rolling)
31:50 Layers of defense: Microsoft Defender for Containers
34:57 Stronger workload isolation with Pod Sandboxing
36:37 Confidential GPU support for sensitive AI workloads
40:08 Preserving state and context in AI lifecycles
43:48 Accelerating inference with Persistent KV Cache
44:48 Scaling value: Cost and efficiency in AKS
49:33 Provisioning requests for Cluster Autoscaler
50:33 Virtual Nodes v2 for bursting and serverless headroom
53:50 Ensuring availability and resilience
57:50 Testing resiliency with Azure Chaos Studio
1:00:10 Early detection, diagnosis, and GPU memory profiling
1:03:00 Cluster Health Monitor
1:08:42 Extending AKS to customer-owned infrastructure
1:12:20 Flex Nodes explained
1:13:17 Demo: Managing a fleet of diverse environments
1:20:00 Recap and concluding thoughts
1:23:38 Invitations to KubeCon and Ignite

[eventID:27395]
Run and Scale AI Applications with Azure Kubernetes Service (AKS)Build an AI Event Planner Agent with Copilot Studio & DataverseFrom Insight to “What If?”: Planning in Microsoft Fabric (Option 2)New Year, New Bot - Build Its AI Brain with .NET and AspireAzure Incident Retrospective: Networking – Gateway service degradation (ID: 7Q30-010)GitHub Learn at GitHub Universe 2026 PreviewFabric Apps in Action: What the Community BuiltFabric IQ in Action: Ask, Understand, ActGuardrail Your Operations with Azure SRE AgentEpisode 3: Prompt Engineering with Microsoft Copilot – Achieving Better Results with AIAzure Retirement Livestream: NSG flow logs in Azure Network Watcher (ID: 0P59-60Z)Working With Your Data, Safely: Files, Approvals and Memory
Microsoft Reactor |

Run and Scale AI Applications with Azure Kubernetes Service (AKS)

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER