product-recommendation-system

AI recommender system with baseline recommendations, data validation, evaluation, FastAPI, and future multimodal RAG-agentic personalization.

A multimodal AI/ML recommendation system that relates and ranks products using three complementary signal types: user behavioral interactions, product text, and product images. It is built on two independent real-world dataset tracks, RetailRocket behavior events and Amazon Berkeley Objects (ABO) product metadata and images, and covers a structured local engineering workflow from dataset discovery and reusable cleaning through baseline and multimodal similarity models, offline evaluation, and a controlled agentic recommendation workflow.