AI / RAG Project – Swiggy Annual Report QA System
Developed a Retrieval-Augmented Generation (RAG) application using Python and Streamlit to answer user queries from the Swiggy Annual Report PDF.
Implemented semantic search using Sentence Transformers (MiniLM) embeddings and cosine similarity to retrieve relevant document chunks.
Integrated GPT-2 from HuggingFace Transformers to generate contextual responses based on retrieved content.
Built an end-to-end pipeline including PDF text extraction, chunking, embedding generation, context retrieval, and answer generation.
Designed an interactive UI displaying answers, similarity scores, and source context for transparency.
Tech Stack: Python, Streamlit, HuggingFace Transformers, SentenceTransformers, PyTorch, Scikit-learn, NLP, RA