AI PDF Question Answering System (RAG + OCR Enabled)

An intelligent system that allows users to have natural language conversations with their PDF documents, leveraging advanced AI to understand content, including text from scanned images.

This project addresses the challenge of extracting valuable information from PDF documents, which often serve as containers for vast amounts of knowledge. The system is designed to intelligently process these files and provide precise answers to user questions posed in natural language. A key innovation is its integration of Optical Character Recognition (OCR), which empowers it to handle not just text-based PDFs, but also scanned documents and images, making previously inaccessible information readily available for querying. The core of the system is a sophisticated Retrieval-Augmented Generation (RAG) pipeline. This process begins by breaking down documents into manageable, semantically-rich chunks. These chunks are then converted into high-dimensional vectors using advanced embedding models and stored in a specialized vector database. When a user asks a question, the system searches this database for the most relevant document segments, ensuring that the context provided to the language model is both accurate and concise, leading to high-quality, relevant answers. To complete the workflow, the retrieved context is passed to a Large Language Model (LLM), which can be run either locally for maximum privacy or via an API for greater power. The LLM then synthesizes the information to generate a human-like answer. The system is highly tunable, with controls for context window size, the number of retrieved chunks (top-k), and token limits, allowing for a fine balance between response speed and detail. The entire process is wrapped in a user-friendly interface, demonstrating a strong command of modern NLP architectures, vector databases, and end-to-end AI system design.