RAG Project - Link to project

Engineered an end-to-end Retrieval-Augmented Generation (RAG) backend using Python and FastAPI, processing PDF documents through text extraction, recursive chunking, local embeddings, semantic retrieval, and LLM-generated responses with source citations. Implemented PostgreSQL + pgvector vector storage and cosine-similarity search with 384-dimensional Sentence Transformer embeddings, enabling context-aware retrieval across uploaded documents without relying on hosted embedding APIs. Integrated a locally hosted Ollama LLM through REST APIs and built a modular, production-oriented architecture with SQLAlchemy, Alembic migrations, Pydantic validation, and a 108-test pytest suite covering ingestion, retrieval, search, and generation workflows.