RAG for Warranty Claims

Built a retrieval-augmented generation pipeline during my internship at Cummins to answer warranty claims questions for engine products.

Pipeline:

• Extracted unstructured warranty claim data from PDF files using document parsing
• Generated embeddings and stored in a vector database for semantic retrieval
• Answer generation using LLMs with retrieved context via Langchain and ChromaDB

Tech: Python, Langchain, ChromaDB, PDF parsing