Demo: AI Study Companion, a RAG study platform built and deployed on AWS

AI Study Companion is a full-stack RAG platform I built as my final year project. It won a bronze medal at my university’s final year exhibition, judged by industry experts. Students upload PDFs, DOCX and PPTX files or paste a website URL, and get answers grounded only in their own material, with citations to the exact file and page. What it does: Document chat with source citations Notes workspace with an AI assistant that has persistent memory Study material studio: summaries, mind maps, flashcards and practice questions (MCQ, short answer, long essay), graded by AI How it works: Ingestion runs in the background (Celery, Redis); Unstructured.io splits documents into text, tables and images, and GPT-4o vision makes tables and images searchable Multi-query hybrid search with Reciprocal Rank Fusion and a Cohere reranker LangGraph agents with LLM-as-judge guardrails for toxicity, prompt injection and PII RAGAS evaluation: 0.92 faithfulness and 0.95 context precision on 100 questions Stack: FastAPI, LangGraph, Next.js, Supabase (pgvector), Docker, AWS ECS Fargate, GitHub Actions Built by Mueed Hyder Mir