About

I’m Michael. I’m a principal engineer with 15+ years building production systems and 4+ years hands-on with generative AI. I build governance infrastructure for LLM-powered applications—systems that enforce deterministic behavior, validate outputs, and make AI safe to operate in production.

LlamaBrain is my open-source framework for this: a 9-component architecture that treats the LLM as an untrusted component with bounded authority. Validation gates, prompt contracts, fail-closed fallbacks, replayable audit traces. 100% constraint success rate across 3,000+ tests.

I specialize in hard problems most teams avoid: determinism under uncertainty, failure-path design, and infrastructure that proves correctness instead of asserting it.

I also maintain Lost Legends, a C89 MUD I’ve run continuously since 1996—30 years of uptime on a legacy codebase I still ship updates to.