Microsoft Reactor Toronto

Scaling LLMs to production introduces critical challenges: How do you orchestrate multi-node execution?

Optimize GPU scheduling? Achieve low-latency data transfers between nodes? This session reveals how NVIDIA Dynamo and Azure Kubernetes Service solve these challenges together.

Through a real-world e-commerce recommendation scenario, discover how open-source frameworks and managed Kubernetes unlock enterprise-grade inference performance.

Learn to harness advanced hardware like GB200 NVL72, optimize distributed workloads on AKS, and deliver cost-efficient AI at scale to turn infrastructure complexity into competitive advantage.

In this session we will build a fully connected .NET 10 solution that will be the core of OpenClawNet. We will wire up .NET Aspire for service orchestration, build the Gateway API using Minimal APIs, add real-time chat with SignalR, create a Blazor web UI, connect to Ollama for local AI inference, and persist chat history with SQLite and EF Core.

At the end we will have a fully functional local chatbot in your browser. Send a message, watch it stream back from your local LLM, and see it persist to the database. A solid foundation for everything that comes next.

Key Topics & Technologies:

• Solution architecture and project organization
• Aspire service orchestration and health checks
• ASP.NET Core Minimal APIs and endpoint design
• SignalR for real-time WebSocket communication
• Blazor (Server-side rendering) for interactive UI
• Ollama integration for local LLM inference
• SQLite with EF Core for session persistence
• Gateway routing and middleware

Pre-requisites:

• Experience: Intermediate C# and ASP.NET Core knowledge (not a .NET 101)
• Environment: .NET 10 SDK, VS Code + GitHub Copilot (or Visual Studio), Docker (for Ollama)
• Optional: Azure subscription (for Sessions 3–4, but local development works fine throughout)
• Mindset: Curiosity about AI agents, willingness to follow along live and ask questions