Fifteen years working with my hands before AI. Electronics and board-level repair (SMD micro-soldering), running retail tech operations, and licensed security work. Different worlds, one habit that carried over: you don't call something fixed until you've proven it fixed.

The last eighteen months, full-time on AI. I design and build AI systems in deep partnership with LLM coding agents — the conductor model: I don't type every line, I direct a team of agents with an engineer's discipline (strict TDD, adversarial review, verification before anything ships). Self-taught, from the primary literature: I read the papers, then implement them.

No computer-science degree — and I lead with that. My case isn't a title; it's open-source systems you can clone and run. A technical diploma, then continuous self-education. I'd rather be judged on the work.

One throughline across everything I build: trust. A memory layer that refuses unsupported facts; a security platform that blind-validates its own findings; a reviewer that attacks its own code. The question is always the same — can you actually rely on what the machine just told you?