System Requirements
- Candidate Generation: Retrieve ~500 candidates from billions in < 20ms
- Ranking Model: Score candidates using feature store & user telemetry
- Near-instantaneous feedback integration when user watches or skips
- Diverse results preventing narrow filter bubbles
Graded Constraints
01Budget within $50,000/month
02Edge CDN and API Gateway for video catalog distribution
03Stream ingestion (Kafka) for view history and interaction telemetry
04Fast feature store / cache (Redis) and Search / Embedding index
05Persistent data lake / DB for watch history and content catalog
Ready to design?
Open this challenge in the interactive system design playground and satisfy all architecture constraints.
Explore More
All ProblemsRelated Architecture Challenges
GoogleEasy
Gmail Spam Filter
Design a real-time email ingestion and spam classification pipeline capable of evaluating incoming messages with ML heuristics, rule-based fallbacks, and user feedback telemetry.
100M+ emails/day →
UberMedium
Scalable Notification System
Architect a multi-channel notification engine supporting iOS APNS, Android FCM, SMS (Twilio), and transactional emails with priority routing and deduplication.
10M+ notifications/day →
NetflixHard
Netflix Home Page Recommendation
Design Netflix's personalized row generation system that curates customized titles, tailored artwork thumbnails, and dynamic row rankings for every member profile.
250M paid subscribers →