In speed skating, tenths of a second can mean the difference between an athlete winning gold or missing out on a medal altogether.
To capture the moment, the U.S. Olympic Speed Skating Team put traditional analysis on the ice and turned to new tools powered by artificial intelligence to simulate the complex aerodynamics of skaters.
The new app, called Slippery Fish, was custom-designed for the team and is part of a new suite of AI-powered tools used by the U.S. Olympic team to improve athlete performance ahead of the 2026 Milan-Cortina Winter Games, which open Friday.
Emery Lehman, 29, a member of the speed skating team competing in this year’s Olympics, said AI-enabled apps have revolutionized the way he and his entire team approach training.
“We used to spend all our time and money flying our players into wind tunnels across the country, where they would stay in a stationary position and allow the team to collect aerodynamic data,” Lehmann told NBC News last week while working out on a stationary bike during a team practice in southern Germany.
“With this app, everything is done through AI,” said Lehman, an engineer who won a bronze medal at the 2022 Beijing Olympics. “For example, in team pursuit, we want to find a position among the three guys or girls skating with modifications so that the whole group can be more aerodynamic.”
The Slippery Fish app is based on a similar AI-powered aerodynamic analysis tool for cyclists called AiRO, which allows coaches to upload images of their athletes on the ice. From these images, the app creates a digital avatar of the athlete and simulates how different positions affect airflow and drag, variables critical to speedskating success.
“We can now incorporate posture changes into this app and see if those adjustments are actually efficient,” Lehman said. “Then you can go on the ice and see if it’s practical and evaluate it from there. What used to take maybe a week or two to validate and say, ‘That was a good idea, that was a bad idea,’ can now be done in a day.”
Chief Shane Dormer Regarding the speed skating team’s sports performance, he agrees that AI-powered technology has improved the way the team trains, calling it a “wind tunnel in your pocket.” “We want to test new positions in a short period of time so we can iterate on the fly, have real conversations and really involve coaches and players in the process,” Dormer said. “We’re looking at small adjustments, such as making sure the skater’s elbows aren’t coming off the body, and what the time costs are.”
AI systems have become increasingly capable over the past year, able to analyze large amounts of data, generate insights, and provide detailed recommendations based on new research techniques. To leverage these growing capabilities, the U.S. bobsled and skeleton team announced a new partnership in November with Snowflake, one of the world’s leading AI-focused data analytics companies.
To director Kurt Tomasevich To improve the sports performance of the bobsled team, this new collaboration provides a deeper understanding of the strengths and weaknesses of individual athletes, allowing coaches to provide better advice and improve the team’s overall performance ahead of the Milan-Cortina Games.
“It seems so easy to put two or four athletes on a sled,” said Tomasevic, who won gold in the four-man bobsled at the 2010 Vancouver Olympics and holds a PhD in bioengineering. “But when you have four athletes running and you ask each athlete to start at different times before they feel like they’re maximizing their push, it sometimes goes against their natural tendencies.”
“Now, what if we could train this AI tool to tell this particular athlete, on this particular day, on this track, how many steps should they take before getting on the sled to reach their optimal speed? That would be amazing.”
Tomasevic said that during training, he only had access to some of the data currently feeding Snowflake’s AI tools.
“In 2006, my first Olympics, we had to split up every few hundred meters on the track,” Tomasevic said of the timing checkpoints. Bobsleds are now equipped with accelerometers and gyroscopes that provide 100 data points per second.
“We’re talking about thousands of times more accuracy. And from a turnaround time standpoint, being able to take the data, upload it to the AI, ask questions, download the results, and give feedback back to the pilot and coach between runs so they can make adjustments before the next run starts.”
Snowflake principal Mike McCarver said the new wave of AI-powered data analysis goes beyond simple AI prompts and is making a big difference to the team’s training routines.
“This really opens up an opportunity for sports teams, athletes, and even brands in general to do more with data that might not have been really digestible otherwise,” McCarver told NBC News.
The Milan-Cortina Games may be the first Olympics to feature extensive use of AI by athletes and coaches, but Dan Webb, director of performance analysis at the U.S. Olympic and Paralympic Committee, said he believes teams are just beginning to explore how AI can transform their work.
“I get questions about AI all the time from different teams,” Webb told NBC News. “While we have made targeted initial investments in some AI applications, we have not yet built an AI tool or suite of AI tools to deploy across Team USA.”
Speedskating’s Dormer believes his team is just beginning to mine AI-derived insights to improve performance and, hopefully, medals at Milan-Cortina and other competitions.
“I think AI will be used more and more as we get through these Olympics,” Dormer said. “We found some crazy things in the software that you can do after the game. It’ll be a lot of fun.”
