video1.0<iframe src="https://www.loom.com/embed/ed55d0b4c60f4cdaa4d67b9c6a2f155a" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/ed55d0b4c60f4cdaa4d67b9c6a2f155a-e8c1f84d4f07d681.gif314.593Hi, this is Karthik. I built an AAPowerDrive chatbot where users upload documents and ask questions, and the system retrieves relevant chunks using a Pinecone vector database and answers with an LLM in a retrieval architecture. I use FastAPI for the backend, Gradio for the UI, and deploy on AWS EC2, with lightweight Lightbit embeddings for AWS free tier. It splits documents into chunks, creates embeddings, and stores them, then returns answers with content scores and the document source. No action was requested from viewers.