AI Autograder

Built a full-stack AI autograder with React (frontend), Django + PostgreSQL (backend), and FastAPI for modular LLM services.Integrated GPT-4 via LangChain and RAG pipelines for rubric-based grading, semantic answer-key matching, and personalized feedback.Applied spaCy, scikit-learn, and sentence-transformers for text parsing, vectorization, and semantic similarity scoring.Used FAISS and Redis for fast vector search and caching, enabling real-time grading and dynamic document retrieval.