Design an Ads Ranking System with Calibrated Predictions

Asked at Meta · System Design for ML · Difficulty: Hard · Last asked May 30, 2026

Design the machine-learning system that decides which ads appear in a social feed. Each time a user scrolls to an ad slot, the system must pick a few ads out of millions, rank them by the value they bring to advertisers, and keep the feed pleasant enough that users keep coming back. The predicted probabilities also set prices in the ad auction, so they must be accurate in absolute terms, not only in order.

Loading the problem…