I build the infrastructure that catches AI agents when they lie, drift, or fail silently. Six trust and verification projects shipped in two months, all Python, all open source, all asking the same question: can you trust what the machine just did?

Agent auditing systems that compare what an AI claims it's doing against what it actually does. Red-team pipelines that stress-test agentic workflows under adversarial conditions. Lightweight CLI tooling for developers who want direct control over agent behavior without platform overhead.

Teams deploying AI agents in production who need verification as a default, not an afterthought. If you're shipping autonomous systems and nobody's checking whether they're doing what they say, that's the problem I solve.

Stack: Python end to end. Philosophy: if you can't audit it, you can't trust it.