For eight years most of what I have learned has lived in design docs, incident reviews and code comments that only my teammates ever read. This blog is an attempt to move some of that thinking into the open.
What to expect here
Three kinds of posts, roughly:
- Infrastructure for AI agents. I work on LLM gateways, multi-agent orchestration and the governance layer underneath enterprise AI. A lot of the interesting problems there — identity for agents, quota design, routing across model providers — are barely written about anywhere. I want to change that, within the limits of what I can share publicly.
- Distributed systems, learned the hard way. The lessons that only show up in production: pipelines under privacy constraints, reliability work nobody notices until it is missing, performance investigations that start with a vague complaint and end somewhere unexpected.
- Working notes. Shorter posts on things I am building or reading — projects, tools, papers, and the occasional opinion I will probably revise.
Ground rules
Everything I publish about employer work stays public-safe and generalised — the problem shapes and the reasoning, never the proprietary detail. Where a post covers team work, I will say which parts were mine.
If any of this sounds useful, the contact page is the best way to tell me what you would like to read about.