VideoRAG: Multi-Modal Video Question Answering

Developed a custom end-to-end RAG pipeline for semantic question answering over video with time-aware

responses, built from scratch without high-level frameworks to maintain full control of retrieval and

generation.