At 9H Digital, I built and shipped 9H Chat in Python and FastAPI, bringing agent conversations, vector retrieval, tool use, and structured output to 50 clients. It handles about 10,000 conversations a month, and I also built backend components and APIs for AI features across enterprise apps serving 2M+ users.
I built evalspec, RAG Probe, and posthog-llm-doctor to make agent evaluation, retrieval quality, and LLM observability measurable. In my work, I use benchmarks, tracing, and documented evaluation results to catch regressions and understand production failures.

