How I Work
Chapter 01
Principles
Data correctness first
I focus on systems where the data has to be right, and I start with the design and the data model because they are the expensive things to get wrong.
End to end
I usually own a change from design through production support. Production is part of the job, not a handoff.
Agent-assisted, independently verified
I’ve built with language models since 2023: internal assistants at work, and local inference, retrieval with citations, and training experiments on my own time. I also build with coding agents every day. Code an agent writes is a change I didn’t write, so it gets independent code review and mutation testing before I trust it.
Leading as a contributor
I’ve managed 12 engineers across 4 project teams and led engineers as a principal. I prefer to set direction while still writing the code.
Chapter 02
Working with AI agents, and leading as a contributor
I've built with language models since 2023, both in internal tools and as part of how I write software. It started out as copying out of ChatGPT. Now agents write most of the code, independent review and tests check all of it, and I read the critical pieces myself. We've still got a long way to go, too.
I've been a lead, a principal engineer, and an engineering manager responsible for 12 engineers across 4 project teams. I'm glad to mentor, review, and set technical direction.