Personalized Content Recommendation System

A Netflix-style hybrid recommender built on MovieLens 25M.

Features

• Content-based retrieval using Sentence Transformers + FAISS
• Collaborative filtering with user-based kNN on implicit feedback
• Hybrid re-ranking combining content and CF signals
• LLM-based explanation generation using T5
• Offline evaluation using Precision@10 and NDCG@10
• Interactive Streamlit dashboard

Tech Stack

Python, PyTorch, FAISS, Hugging Face Transformers, Streamlit

How to Run

pip install -r requirements.txt
streamlit run app.py