Security engineer building tools that make application security work faster without making it careless. Currently building VIPER, submitted to OpenAI Build Week 2026. Codex and GPT-5.6 were used as a real engineering partner this Build Week: tracing callers, capturing real network traffic to diagnose broken routes, writing adversarial regressions, and reviewing design decisions before anything shipped.

A few things I actually believe: proof before confidence, quality before coverage, and a gate with 1,635 passing tests and three failures is still a failure.

A bounded, evidence first control plane for application security research, with a live public demo.

_{Currently caffeinated. Ask me about the time I spent an hour debugging a demo before realizing GitHub Codespaces itself was the bug.}