Build or Fix a RAG / LLM Application by Sachin RajanBuild or Fix a RAG / LLM Application by Sachin Rajan

Build or Fix a RAG / LLM ApplicationSachin Rajan

Cover image for Build or Fix a RAG / LLM Application

I build, debug, and improve RAG and LLM applications with a focus on retrieval quality, reliability, and production-ready backend implementation.

If your current system retrieves the wrong context, produces poorly grounded answers, struggles with document ingestion, or is still stuck at prototype stage, I can work on the full pipeline rather than only tweaking prompts.

Typical work includes:

• Document and data ingestion pipelines • Parsing, cleaning, chunking, and metadata design • Embedding generation • Vector database integration • Semantic search and metadata filtering • Hybrid retrieval • Reranking • Prompt and context construction • LLM API integration • Structured outputs • Citation and source handling • Retrieval debugging and evaluation • FastAPI endpoints and backend integration • Improving existing RAG systems

What you’ll receive:

• Working RAG / LLM pipeline • Ingestion and retrieval implementation • Vector search integration • LLM integration • Retrieval quality improvements • Error handling and validation • Testing of key workflows • Setup and handoff documentation

I can build a new system from scratch or work inside an existing codebase.

My approach is retrieval-first. Good RAG is not just a prompt plus a vector database. The quality of the final answer depends on ingestion, chunking, metadata, retrieval, reranking, context construction, evaluation, and the model working together as one system.

Example work