The NBA is leveraging AI to process data on players' live body movements, analyzing metrics such as speed, dunk height, number of passes, dribbles, and injury risk in real time. This is achieved through a Microsoft Azure solution built on Azure Kubernetes Service (AKS), which can manage and process up to 16 gigabytes of raw data per game. The insights gained help teams understand player strengths and weaknesses, ultimately improving performance.

The NBA Stats Team collects millions of data points for each game, focusing on 29 specific points on each player's body. This data is processed using a computer vision-based player tracking system, which operates at 60 hertz, allowing for real-time analysis. The information is stored in Azure Cosmos DB for efficient retrieval and is used to generate insights that can be shared with coaches, broadcasters, and fans.

The NBA intends to build more sophisticated AI models using Azure AI infrastructure as they become more familiar with the technology and data sources. This includes plans to utilize GPU-accelerated virtual machines for enhanced machine learning capabilities. The goal is to improve data search efficiency and potentially provide fans with access to these insights during games.