The Question
ML DesignScalable Personalized Content Discovery System
Design a high-scale, end-to-end recommendation engine for a news or article platform. The system must handle millions of items and users, optimize for diverse engagement metrics (CTR and Dwell Time), and ensure low-latency serving while addressing common challenges like content freshness, cold start, and position bias.
Two-Tower
DeepFM
ANN
Faiss
Kafka
Flink
Spark
Feast
S3
XGBoost
AUC
NDCG