From GitHub to GPU: Build an Open Source AI App with Copilot, Azure Linux, and AKS @MicrosoftReactor
From GitHub to GPU: Build an Open Source AI App with Copilot, Azure Linux, and AKS  @MicrosoftReactor
Uploaded October 2026 | Updated November 2026, 1 day ago
What does it take to turn an AI idea into a real, scalable application without hiding the infrastructure behind a black box?

We'll start with an open source GitHub repository and use GitHub Copilot to understand, adapt, and operate a complete AI workload. We deploy Ray and KubeRay on Azure Kubernetes Service (AKS), run LLM batch inference on an NVIDIA GPU, and explore how Azure Linux provides a lightweight, secure container host for the Kubernetes nodes under the hood.

Along the way, we'll connect the layers that make the application work using containers, Kubernetes resources, GPU scheduling, and open source operators.

We'll also explore autoscaling, network isolation, authentication, metrics, dashboards, and cost-aware management. Leave with a reusable GitHub repo and a clear roadmap for building open source AI systems on Azure.

🔗 Learn more by exploring the resources: https://aka.ms/MicrosoftAzureOpenSourceBlog

📌 This event is part of a series. Learn more here: https://aka.ms/AIwithOpenSourceSeries/y

0:00 Introduction and housekeeping
0:50 Welcome and overview of the AI stack
3:14 Project repository and dashboard introduction
5:34 How GitHub Copilot was used to build the app
14:16 What is Ray and why use it on Kubernetes?
16:37 Intro to KubeRay and GPU scheduling
19:12 Building the platform with AKS and T4 GPUs
23:25 Anatomy of a GPU Ray cluster
26:22 Running GPU tests: Interactive vs. Declarative
27:54 Demo: VS Code dashboard and live inference
34:25 Understanding autoscaling layers
40:50 Metrics, observability, and dashboard breakdown
43:40 Network security and access policies
46:27 Authentication with cluster tokens
48:26 Gotchas and challenges from the POC
49:26 Cost awareness and billing tips
52:05 Summary of known issues and roadmap
55:12 Final roadmap to production

#MicrosoftReactor #LearnConnectBuild #OpenSource #Linux #Kubernetes #AIapps

[eventID:27664]
From GitHub to GPU: Build an Open Source AI App with Copilot, Azure Linux, and AKSAI @Work: Community Builders in ActionCopilot como tu primer asistente de IA: diseña un asistente para una tarea realModel Mondays - Spotlight On Fine-Tuned ModelsEpisode 2: Generative AI – Creating, Innovating, and Accelerating with AIEpisode 1: AI Fundamentals – Understanding the Future of WorkCapacidades avanzadas de MCP: Apps, Tasks y WebMCPSkill up on Microsoft Copilot StudioHow to trust the code you shipDe las ideas a la acción con Copilot: convierte una idea en un experimento.NET + Foundry, better togetherAI Agents That Know Your Business (Option 1)
Microsoft Reactor |

From GitHub to GPU: Build an Open Source AI App with Copilot, Azure Linux, and AKS

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER