Design a scalable recommendation serving system

Assume you are designing the request-serving tier for a high-traffic recommendation platform, such as a delivery app or an online marketplace feed. The discussion should be infrastructure-oriented; model design details are not needed. Requirements: When a user opens the app or refreshes their feed, the system must return the K most relevant items. Latency targets: median response time below 50 ms, and the 99th percentile below 200 ms, consistent with standard mobile latency…