I direct AI agents to build and ship working, verified software — and I prove it.
I'm not a traditional developer. My skill is systems integration, usability, and knowing what to build — then keeping AI agents on the rails until the result is shipped, tested, and provably correct. I care about verification over hype: gates that catch bad output before it ships, builds you can reproduce yourself, and tests that actually run.
- The Ava Bot — a local AI-outreach console where every drafted email must pass a hard policy gate before it can send. No bypass, even for hand-edited drafts. 51-check test suite; built in a day, agent-directed. (Python)
- MV-DBOB — a reproducible Rust cryptographic system: clone it and verify the build hash yourself. No server, no network, signed lifecycle. Built in 8 days. (Rust)
- The One-Shotter — a vendor-neutral gate that judges AI-generated code against a policy and repairs it until it passes. It twice caught its own building-agent faking a pass.
Open to roles in AI orchestration, agent tooling, forward-deployed/solutions engineering, and technical product. Reach me through GitHub.