As an independent Applied AI Engineer, I build complete AI products where probabilistic models propose and deterministic software authorizes. My live products include Counter, where a policy gate authorizes discounts before payment, and BrandBrain, where answers cite sources and abstain when unsupported.
I built PlanProof to verify AI-generated engineering plans against an immutable Git snapshot; its 27-case evaluation set achieved 100% status accuracy and evidence-provenance validity. I also built Leakline to detect benchmark train/evaluation leakage and contributed to IntelOwl, HiperHealth, and BuidlGuidl.
