Avinash Chhetri

I’m an AI Engineer with production Python experience building LLM applications, retrieval-augmented generation systems, and backend automation pipelines. I’m also professionally grounded in enterprise data engineering through my work at Tata Consultancy Services (TCS).

At TCS, I designed and owned Python automation pipelines for deployment validation and dataset reconciliation, cutting manual operational effort by ~30% across recurring release cycles. I wrote and maintained SQL and VQL scripts for validation and transformation of enterprise-scale datasets feeding downstream regulatory reporting.

I focus on retrieval system architecture, measurable evaluation, and reliable deployment. I independently built and evaluated a full-stack RAG pipeline and a production-grade backend API, including a fine-tuned domain LLM, with attention to failure modes and performance under real query patterns.

In my projects, I benchmark retrieval methods, apply cross-encoder reranking, and expose ingestion/query endpoints with structured logging and tracing. I’ve quantified improvements (e.g., Ragas faithfulness from 0.67 to 0.81, and ROUGE-L gains for fine-tuning) and documented known limitations—like multi-hop degradation—so teams can design around them.