Online commercial app marketplaces serve millions of apps to billions of users in an efficient manner. Bandit optimization algorithms are used to ensure that the recommendations are relevant, and converge to the best performing content over time. However, directly applying bandits to real-world systems, where the catalog of items is dynamic and continuously refreshed, is not straightforward. One of the challenges we face is the non-trivial computation costs for large scale systems, which is further aggravated by user privacy related constraints for server side computation. To address this problem we introduce an efficient two-layer bandit approach which is contextualized to user cohorts of similar taste. We mitigate cannibalization at runtime within a single multi-intent content surfacing platform by formalizing relevant offline evaluation metrics, and by involving the cross-component interactions in the bandit rewards. The framework allows flexibility for tradeoffs between compute, storage and accuracy of models. The user engagement in our proposed system has more than doubled as measured by online A/B testings.
Puja Das
Dr. Puja Das, leads the Personalization team at Warner Brothers Discovery (WBD) which includes offerings on Max, HBO, Discovery+ and many more.
Prior to WBD, she led a team of Applied ML researchers at Apple, who focused on building large scale recommendation systems to serve personalized content on the App Store, Arcade and Apple Books. Her areas of expertise include user modeling, content modeling, recommendation systems, multi-task learning, sequential learning and online convex optimization. She also led the Ads prediction team at Twitter (now X), where she focused on relevance modeling to improve App Ads personalization and monetization across all of Twitter surfaces.
She obtained her Ph.D from University of Minnesota in Machine Learning, where the focus of her dissertation was online learning algorithms, which work on streaming data. Her dissertation was the recipient of the prestigious IBM Ph D. Fellowship Award.
She is active in the research community and part of the program committee at ML and recommendation system conferences. Shas mentored several undergrad and grad students and participated in various round table discussions through Grace Hopper Conference, Women in Machine Learning Program colocated with NeurIPS, AAAI and Computing Research Association- Women’s chapter.