Design Video Recommendations

Recommendation systems are the most common ML system design question, and video recommendations are their canonical form: a billion-item catalog, a hard latency budget, rich behavioral data, and a business that lives or dies by the quality of the ranking. This walkthrough follows the Interview Framework end to end for the home feed of a large video platform.

This is the ML companion to our system-design treatment of YouTube; that page covers storage and delivery, this one covers the models that decide what to show.