Show Notes
We start by weighing the trade-offs: managed AI gives you speed, safety, and a deep model catalog, but steady high-volume workloads, strict compliance, or edge latency often tilt the equation. That’s where AKS shines. With managed GPU node pools, NVIDIA drivers and operators handled for you, and Multi-Instance GPU to prevent noisy neighbours, you get reliable performance and better utilisation. Auto-provisioning brings GPU capacity online when traffic surges, and smart scheduling keeps pods where they need to be.
The breakthrough is Kaito, the Kubernetes AI Toolchain Operator that treats models as first-class cloud native apps. Using concise YAML, we containerise models, select presets that optimise vLLM, and expose an OpenAI-compatible endpoint so existing clients work by changing only the URL. We walk through a demo that labels GPU nodes, deploys a model, serves it via vLLM, and validates responses from a simple chat UI and a Python client. Tool calling and MCP fit neatly into this setup, allowing private integrations with internal APIs while keeping data in your environment. Text Us About the Show
More from Microsoft Community Insights Podcast
In this episode chat with the one of the creator behind Gtihub Agentic workflows to find out what inspired him and tell his story behind it. Peli traces the idea from continuous AI with GitHub Next to early prototypes running coding agents in CI/CD, and explains why automation experience from .NET testing and MakeCode learning sandboxes shaped the design. He defines agentic workflows as markdown-defined workers compiled to Actions, with layered security including containers, microVMs, firewalls, MCP-gated access, and a zero-secret policy.
I joined Art in a conversation Microsoft Office 365 copilot adoption and how he used copilot co-work in daily to do daily task. AI Community Conference HK – September 4 (Co-hosted by our APAC team) Boston Data and AI Saturday 2026 – October 3 Community Summit NA 2026 – Nashville, TN – October 12
We talk with Navyasree Potluri who is from Microsoft about FabricIQ Integration
In this episode we discuss the new Foundry Toolbox with Arafat Tehsin. Toolbox is a new way to centralise, configure, and reuse tools across all of your AI agents without re-configuring them in Microsoft Foundry. More information can be found below including link to his demohttps://github.com/arafattehsin/foundry-toolbox-demoshttps://arafattehsin.com/bloghttps://fedorai.com https://devblogs.microsoft.com/foundry/introducing-toolboxes-in-foundry/
Bernat Agulló is a data platform engineer based in Barcelona who speaks Spanish, English, Japanese, and some German. Agulló has been a Microsoft MVP since 2022 and recently renewed the status for a fifth year. His professional background in data engineering began around 2018, following earlier experience with Excel macros, SQL, and data pipelines like SSIS. He gained recognition in the community by focusing on niche topics such as calculation groups and scripting, areas that were not widely covered by other presenters at the time.