Design a product recommendation system for Target.com that can handle millions of users and a vast product catalog, considering real-time recommendations based on the current shopping session.
FR:
Provide recommendation based on user shopping/watch, card, browsing history Recommendation should be provided based on recent activity as well in real time within Sec. Provide Millions of concurrent users and vast product catalog Support multichannel recommendations web, device
NFR
Scale horizontally to handle high request volume and large data sets. Real-time recommendations must be computed quickly (sub-second response times). The system should tolerate node failures and maintain uptime. Ensure recommendations reflect the current session and recent user behavior.
What does user wants as recommendations? how do we scale it?