Question
Design a personalization cold-start system for a marketplace homepage. The hard part is the cold-start edges: brand-new users (no history) and brand-new items (no interactions) must still get reasonable, relevant recommendations from their first session, while established users get fully personalized results. Scale is ~10M daily users and a catalog of ~5M items with thousands of new items added daily. Homepage must render in ~150ms, and you need both new users and new items to start accumulating useful signal quickly rather than being stuck in a chicken-and-egg trap.
Clarify scale and constraints first. Propose a clean component breakdown, then go deep on the hard parts — data model, bottlenecks, consistency, failure modes — and name the trade-offs you are making.