Hello, world

This is the first post on the site. A placeholder mostly, to confirm the rendering pipeline works end-to-end.

What this blog is for

I write about AI engineering — the unglamorous parts. Building RAG systems that don't hallucinate confidently, agents that recover from failure, evals that actually catch regressions. Less about which model came out this week, more about what it takes to ship something that works in production.

What you can expect

Roughly one post a month, longer than a tweet and shorter than a paper. Each post is something I learned the hard way or wish someone had told me earlier.

If you find something useful, tell me. If you find something wrong, definitely tell me.