For many years, the main goal of the Netflix personalized recommendation system has been to get the right titles in front each of our members at the right time. But the job of recommendation does not end there. The homepage should be able to convey to the member enough evidence of why this is a good title for her, especially for shows that the member has never heard of.
One way to address this challenge is to personalize the way we portray the titles on our service. Our image personalization engine is driven by online learning and contextual bandits. Like many other Netflix machine learning algorithms, it started as a prototype and needed to transition into reliable production jobs outfitted with monitoring, alerts, model checking, retraining, detecting stale models, system resiliency, and more.
We will discuss how we approach making machine learning systems at Netflix ready for production and how we scaled our artwork personalization engine to reliably handle over 20 million personalized image requests per second.
Tony is Director of Machine Learning at Netflix and is sabbatical professor at Columbia University. He served as general chair of the 2017 International Conference on Machine Learning. He has published over 100 scientific articles in the field of machine learning and has received several best paper awards. He has co-founded and advised multiple AI startups.