Practical AI that ships and holds up. I design and build LLM features, AI agents, MCP integrations, and RAG systems — with eval harnesses so outputs are measured and traceable, not hype. I've built a private multi-model inference + agent control plane and an AI systems-audit engine with its own accuracy harness. Where most "AI developers" hand you a demo, I hand you something production-grade, governed, and testable. Fixed scope, fixed price, weekly working demos. Portfolio: descrylabs.com.