What I do

Turn messy problems into systems people can build and operate.

Most of my career has involved architecture: understanding the real problem, finding the constraints that matter, making decisions explicit, and giving delivery teams a structure they can execute against.

That same instinct is what pulled me into AI engineering. LLMs are remarkably good at some parts of the job and remarkably unreliable at others. I am interested in designing the surrounding system so we can use the strengths without pretending the weaknesses are not there.

This site

A place to publish what survives contact with reality.

I write when an idea has become useful enough to explain clearly. Some of it comes from successful work. A lot of it comes from things that failed first.

The aim is simple: publish the lessons that are useful beyond the project that produced them.