Case Study

A multi-channel retailer's personalization engine ran on stale, daily-batch customer segments, missing intent signals from same-day browsing and cart behavior.

Built a real-time feature pipeline on Kafka feeding a recommendation model, with a RAG-based support assistant layered on top of the same customer data platform for service teams.

• Kafka-based event streaming for clickstream and transaction data
• Feature store for real-time and batch feature parity
• RAG pipeline grounded in product catalog and order history
• MLOps pipeline for continuous model evaluation