I build systems thatship, not demos thatlook like they might.

I write backends. Django, FastAPI, Postgres, Redis, Celery, Docker — not because the list looks good on a resume, but because that’s the toolchain behind every system I’ve put in front of real users. At Mayflower Women’s Hospital I built the CRM and HR system that manages 1,500+ employee records; it’s boring software in the best sense, the kind people rely on without thinking about it.

The AI work is the same discipline pointed at harder problems. Chitraksha is a bilingual RAG companion for student mental wellness — 500+ documents through Docling and Mistral OCR, FAISS retrieval, Phi-3.5-mini generation, answers in under two seconds. The part no tutorial covers: my users write in Hindi, English, and Hinglish inside the same sentence, and retrieval has to hold up anyway. Over a hundred students use it. Some of them on bad days.

The through-line is the distance between a notebook and a person. Anyone can get a model to work in a demo. Getting it to work at a shooting range over WebRTC while forty athletes wait for their scores — that’s latency, retrieval quality, deployment, and all the unglamorous engineering that decides whether something actually ships. That’s the part I’m good at, and the part I like.