Grishman Paruchuru.

I spend my days at the intersection of large language models, retrieval systems, and the real product surfaces people actually use. I've built RAG pipelines that retrieve the right context the first time, intelligent document processors that turn messy PDFs into clean structured data, and the SaaS platforms that put those capabilities into customers' hands.

The work I'm proudest of doesn't sit in a Jupyter notebook — it ships, it stays up, and it earns trust. That means thinking about evals, latency, cost, retries, schema validation, observability, and UX with the same seriousness as the model itself.

The Playground is a small, runnable retrieval system whose corpus is my own résumé, projects, and notes. Type a question and a deterministic local sim shows you which documents it pulled, the similarity scores, and a grounded answer — the same loop I build at work, just shrunk down for you to poke at.