I build infrastructure for agentic AI systems with a focus on auditable execution, provenance, and fail-closed controls at the action boundary.