Scaling Geo-Temporal ML - How Pokémon Go Optimizes Global Gameplay With Kubernetes and Kubeflow

Discover how Niantic's machine learning team tackles the complex challenge of optimizing Pokémon Go's global gameplay through geo-temporal ML systems in this 15-minute keynote presentation. Learn about the unique machine learning challenges involved in enhancing location-based gameplay for millions of worldwide players using massive volumes of geo-temporal data. Explore the team's approach to optimizing Raid Battle spawns across location, time, and difficulty dimensions at global scale spanning millions of S2 cells, while maintaining high standards for player satisfaction, scalability, and privacy protection. Gain insights into designing geo-aware recommender systems and operationalizing them in cloud-native environments using Kubernetes and Kubeflow. Understand modeling techniques for recommendations across spatial, temporal, and difficulty axes, geo-temporal feature engineering at scale, and building scalable recommender systems. Discover MLOps lessons learned from one of the world's most popular location-based games, presented by Niantic's Director of Big Data and Machine Learning and Staff Machine Learning Scientist at the CNCF conference.