I · Notes from a working desk
Most of my work sits between backend systems, data platforms, and applied AI. Lately, I've been building[1] Axiom Engine — an open-source RAG service focused on quote-backed answers. The interesting half of it is the verifier: catching fabricated citations before they make it back to the user.
Day job is data engineering at[2] PG&E, and before that I was an SDE at[3] Amazon. Different stacks, same shape: software that has to keep working when the data is dirty, requirements move mid-quarter, and the people using the tool can't pause for your refactor.
What pulls me in is the seam where a real user with real constraints meets data that has to be trusted. Take any one of those out and the result tends to be brittle in ways nobody enjoys maintaining. That's the part of applied AI I find worth doing — where the engineering actually lives.
Most software problems aren't really technical. They're somebody's workflow in a technical disguise.
The engineering gets a lot clearer once you look at the workflow first.