Mikhail

Generalist by necessity: Python, MCP protocol design, LanceDB/vector search, AI-agent orchestration. Not a classically trained programmer — I direct, verify, and debug AI-written code.

RAG and retrieval architectures for code (PageRank vs BM25, graph-based context), LLM agent evaluation methodology, and Rust for potential MCP tooling.

Discussing AI-agent-driven development workflows, codebase intelligence / RAG for code, honest benchmarking practices, and MCP server design. Say hey if you've hit similar retrieval or eval bugs.