About
Hi, I'm Slav Kurochkin.
I'm an engineer focused on AI quality and evaluation. I care about one question: how do you know your AI system works, and how do you know it still works after the next change?
I've spent 10+ years building test automation, test infrastructure, CI/CD, and observability, most recently leading developer experience and quality work at Worth AI, and before that at Zillow and LTK. Along the way I've shipped AI agents, MCP integrations, and quality gates that made engineering teams measurably faster and more reliable.
I build the quality stack for AI in production: eval datasets sized so results mean something, LLM judges calibrated against human labels, agent evals that grade trajectories as well as answers, observability that shows what the system actually does, deterministic test automation for everything around the model, and CI gates that catch regressions before users do. I also cover voice AI, including speech-to-text, text-to-speech, and voice agents.
I share what I learn openly through hands-on labs, a YouTube series, and this blog. I'm based in Miami / Fort Lauderdale, FL and work remotely.
Open source
- AI-Assisted Engineering Framework: An operating model for building real software with an orchestrator and specialist AI agents, optimized for reliability over demos.
- mcp-chat: A terminal chat CLI over the Anthropic API with MCP server integration: tools, document resources, and prompt commands.
- signal-agent: A personal signal-filter and Q&A agent for software and AI engineering content.