Roman Romanov

I build AI systems — and write about the messy middle between a research demo and something that actually ships.

My work sits where language models meet real software: tool-calling and the Model Context Protocol, agent architectures, and the plumbing in Python that turns a clever prompt into a dependable service. This blog is where I think that out loud — the failure modes, the design calls, and the unglamorous parts that decide whether a system holds up in production.

🇺🇦 Based in Ukraine.