Design Popular Items Page

Design a system that powers a "Popular Items" page for Uber’s in-app marketplace (e.g., the list of top-selling or most-viewed items shown when a user opens the Uber Eats or grocery tab). The page must return the current top-K items ranked by a composite popularity score that combines real-time order counts, view counts, ratings, and a time-decay factor so older events weigh less. The list should update at near-real-time (tens-of-seconds latency) and be available globally with <200 ms end-to-end response time. Support multiple sliding windows (last hour, today, this week, all-time) and allow business users to configure different popularity formulas per city or vertical (food, convenience, alcohol). The system must handle 5 M events per minute (orders, views, ratings) worldwide, store catalog metadata for ~50 M items, and serve 100 K requests per second from mobile clients. Ensure freshness, fault tolerance, and the ability to inject editorial or sponsored placements into the final ranked list.