System design interview question
Design a Distributed Recommendation Engine and Personalization Service like Netflix's Content Recommendation System at Scale
Design a real-time personalization and recommendation service that serves 200M users globally with sub-200ms latency for collaborative filtering and content ranking. The system must ingest 500K events/second (user views, clicks, ratings), maintain user-item interaction matrices, compute personalized rankings across 10M+ content items, and handle regional preference variations across 5 geographic regions. Key constraints include: strict latency SLAs (p99 < 200ms), eventual consistency acceptable for recommendations (not critical data), high throughput writes from client events, and the ability to A/B test multiple ranking algorithms simultaneously. Address how you would partition user and item data, handle stale or missing user profiles, scale feature computation, manage model serving infrastructure, and ensure system resilience during regional outages or traffic spikes.
Practice this question step by step with an AI interviewer that reviews your requirements, your architecture and your trade-offs, and scores the design.