I design, build and operate production-grade AI systems: Generative AI, LLM applications, autonomous agents and RAG architectures. 5+ years of taking AI research and turning it into working enterprise products.
Narau (open source, coming soon) is a long-lived, self-learning agent: the longer it lives, the smarter it gets. It learns from your corrections, crystallizes successful work into new skills, and grows into a personal AI twin — your assistant that writes and acts in your voice. Domain abilities ship as packs (content avatar, research digest, code assistant), so one core serves many jobs.
Every learned change passes a governed pipeline: adversarial review, an objective eval gate and human approval before it is adopted — with a signed, tamper-evident history of every decision.
Self-improving agents are a supply-chain risk: every skill they install is somebody else's code. Narau's answer is verification, not volume — self-improvement you can trust.
GenAI & Agents:
ML:
Languages:
Platform, Data & Cloud:
