Team USA’s skiers and snowboarders are going home with some new hardware, including a few gold medals, from the 2026 Olympics. Along with the years of hard work that go into being an Olympic athlete, this year’s crew had an extra edge in their training thanks to a custom AI tool from Google Cloud.
US Ski and Snowboard, the governing body for the US national teams, oversees the training of the best skiers and snowboarders in the country to prepare them for big events, such as national championships and the Olympics. The organization partnered with Google Cloud to build an AI tool to offer more insight into how athletes are training and performing on the slopes.
Video review is a big part of winter sports training. A coach will literally stand on the sidelines recording an athlete’s run, then review the footage with them afterward to spot errors. But this process is somewhat dated, Anouk Patty, chief of sport at US Ski and Snowboard, told me. That’s where Google came in, bringing new AI-powered data insights to the training process.
Google Cloud engineers hit the slopes with the skiers and snowboarders to understand how to build an actually useful AI model for athletic training. They used video footage as the base of the currently unnamed AI tool. Gemini did a frame-by-frame analysis of the video, which was then fed into spatial intelligence models from Google DeepMind. Those models were able to take the 2D rendering of the athlete from the video and transform it into a 3D skeleton of an athlete as they contort and twist on runs.
The AI model running on the screen in the background shows how the tool tracks an athlete’s performance.
Final touches from Gemini help the AI tool analyze the physics in the pixels, according to Ravi Rajamani, global head of Google’s AI Blackbelt team. which worked on the project. Coaches and athletes told the engineers the specific metrics they wanted to track — speed, rotation, trajectory — and the Google engineers coded the model to make it easy to monitor them and compare between different videos. There’s also a chat interface to ask Gemini questions about performance.
“From just a video, we are actually able to recreate it in 3D, so you don’t need expensive equipment, [like] sensors, that get in the way of an athlete performing,” Rajamani said.
Coaches are undeniably the experts on the mountain, but the AI can act as a kind of gut check. The data can help confirm or deny what coaches are seeing and give them extra insight into the specifics of each athlete’s performance. It can catch things that humans would struggle to see with the naked eye or in poor video quality, like where an athlete was looking while doing a trick and the exact speed and angle of a rotation.
“It’s data that they wouldn’t otherwise have,” Patty said. The 3D skeleton is especially helpful because it makes it easier to see movement obscured by the puffy jackets and pants athletes wear, she said.
For elite athletes in skiing and snowboarding, making small adjustments can mean the difference between a gold medal and no medal at all. Technological advances in training are meant to help athletes get every available tool for improvement.
“You’re always trying to find that 1% that can make the difference for an athlete to get them on the podium or to win,” Patty said. It can also democratize coaching. “It’s a way for every coach who’s out there in a club working with young athletes to have that level of understanding of what an athlete should do that the national team athletes have.”
For Google, this purpose-built AI tool is “the tip of the iceberg,” Rajamani said. There are a lot of potential future use cases, including expanding the base model to be customized to other sports. It also lays the foundation for work in sports medicine, physical therapy, robotics and ergonomics — disciplines where understanding body positioning is important. But for now, there’s satisfaction in knowing the AI was built to actually help real athletes.
“This was not a case of tech engineers building something in the lab and handing it over,” Rajamani said. “This is a real-world problem that we are solving. For us, the motivation was building a tool that provides a true competitive advantage for our athletes.”
