The modern F1 pit crew: AI engineers who race the deadline, not the car

AI vendors are starting to send engineers into customer operations, and Formula One is a brutal place to try the idea. The Aston Martin Aramco Formula One Team now has CoreWeave’s physical AI engineers working alongside its own staff, in a sport where races come every two weeks and the difference between good and bad can be measured in milliseconds.

The attraction is not just racing know-how. Richard Ahlfeld, CoreWeave’s senior vice president of physical AI, said the company’s real edge comes from years spent helping leading AI companies set up infrastructure and run models. That experience, he argued, teaches what actually matters when systems have to work under pressure.

Eric Ernst, the team’s commercial technology ambassador, said each car generates a million data points per second. That leaves very little room between final practice and qualifying to turn raw information into setup calls. External engineers help because the schedule never waits, and neither does the development cycle. “You got to hit a deadline,” Ernst said.

One of the first projects focused on team radio. Across the grid, there are dozens of open channels, and people have traditionally had to listen in manually. CoreWeave’s model now listens to 40 people at once, writes a summary, and already knows what to do from a prompt. If a car is pitting, it can trigger the next steps within 30 seconds.

The team also taught the system the language of racing, so terms like “oversteering” and “overheating” stay distinct even with engine noise in the mix. And CoreWeave says that’s becoming its broader field engineering playbook: more than 100 projects, all starting from a customer’s deadline and working backward. The lesson is refreshingly unsentimental. Not every problem needs AI. Sometimes a rule is just a rule.