Retail-GPT: Hyper-Personalized Offer Engine

Retail-GPT is a hyper-personalized offer engine that leverages Redis as a unified AI memory layer. It utilizes Redis for vector search with HNSW indexing, semantic caching to reduce AI call costs by over 60%, and Redis Streams for real-time event processing. The architecture integrates vectors, cache, features, and session state within Redis, enabling sub-50ms latency for personalized offers. The project demonstrates innovation beyond simple caching by enabling AI-powered, context-aware offer selection and real-time learning from user interactions. Performance targets are met, with median latency around 35ms and a cache hit rate of approximately 70%. The submission highlights Redis Cloud's scalability and readiness for production environments, addressing real retail personalization challenges with a clear ROI. Retail-GPT positions Redis as the essential brain for modern AI applications, particularly in real-time personalization. The demo is easily runnable with a single command, showcasing all Redis AI capabilities. This project represents a final submission for the Redis AI Challenge 2025.