Scalable MLOPS for in-game AI Features: From highlight detection to player behavior modeling

The integration of artificial intelligence (AI) in modern gaming has enabled dynamic and personalized in-game experiences, including real-time highlight detection and adaptive player behavior modeling. Central to operationalizing these AI features is the application of machine learning operations (MLOPS)—a framework that streamlines model development, deployment, and monitoring at scale. This review synthesizes current methodologies across deep learning, reinforcement learning, and imitation learning in the gaming context, highlighting the role of MLOPS in ensuring system robustness and scalability. Experimental results show the superiority of transformer architectures for highlight detection and behavior cloning methods for imitation learning. We also discuss operational bottlenecks, ethical considerations, and propose future directions including meta-learning, federated training, and energy-efficient AI infrastructures. This paper aims to serve as a comprehensive reference for researchers and practitioners in gaming AI and scalable MLOPS systems.