This document summarizes machine learning challenges at Criteo, an online advertising company. It discusses how Criteo uses machine learning for bidding, product recommendations, and banner selection. It also outlines some of Criteo's machine learning challenges, including optimal bidding strategies under uncertainty, probabilistic cross-device matching, and modeling long tail users and products. The document concludes that while machine learning applies well to online advertising at scale, there is still room to improve the user experience through new algorithms and making sense of new data sources.