Umangi Nigam

At Yellow.ai, I maintain production ML and Generative AI pipelines across preprocessing, training, evaluation, and inference, improving response accuracy by 30%. I also work directly with technical customers and client engineering teams to integrate AI workflows into live infrastructure.

I've architected AWS S3 and auto-scaling Lambda ingestion and indexing pipelines for 10K+ documents, and built FastAPI APIs, vector retrieval systems, and production AI automation that reduced query response times by 40%.

I build practical LLM products including an AI cofounder with persistent memory, a real-time AI narration platform, and a token optimization engine that cut repetitive-traffic token use by up to 80% while measuring quality trade-offs.