This newsletter is where I share practical notes on building reliable AI-enabled systems with .NET + Azure — the kind of stuff that helps you move from “it works on my machine” to “it works in production”:
- Shipping patterns for AI features in .NET (APIs, background jobs, integration boundaries)
- Reliability & debugging (observability, timeouts, failure modes, cost surprises)
- Lightweight quality gates that catch regressions early (tests, baselines, guardrails)
- Short field notes from experiments (local models, vector search, tooling)
- Occasional deep dives when something is worth a full write-up
If you’re new, start here:
- Local LLMs in .NET (Ollama + local workflow)
https://www.lukaswalter.dev/posts/local-llms-in-.net/
- Debugging LLM timeouts in .NET (repeatable triage)
https://www.lukaswalter.dev/posts/debugging-llm-timeouts-in-.net/
- Eval-first: Why “It Worked Once” Is Not a Sign of Quality
https://www.lukaswalter.dev/posts/eval-first-in-.net/
What I’m working on:
- A first public version of my Qdrant + .NET repo (getting started + best practices)
P.S. In case you missed it: Microsoft’s Agent Framework reaches Release Candidate https://devblogs.microsoft.com/foundry/microsoft-agent-framework-reaches-release-candidate/