I design and ship production-grade AI systems — not proof-of-concepts that stay in a notebook. My work spans the full lifecycle: architecting a RAG pipeline, choosing the right retrieval/vector strategy, wiring it into automation (n8n) and real APIs, and hardening it enough to run unattended in production.

I care about the boring parts that make AI systems reliable: sane fallbacks when a model call fails, cost-aware architecture, and clear boundaries between the LLM layer and the rest of the system — the same engineering discipline as any other backend service, applied to GenAI.

mindmapAI — an LLM-backed tool that converts unstructured notes into structured mind maps — is one example of that end-to-end approach.