The first Olympic champion built on her own AI? Kristen Faulkner is rewriting the rules of performance

The American Olympic road race champion revealed via an update on Linkedin that she has spent the past two months living a double life. Training at the highest level of professional cycling by day, and writing code around her training blocks, often for hours at a stretch. Rides end, laptop opens. Sometimes before her bike is even put away.

What pushed her there were both curiosity and frustration. The answers she needed about her own physiology simply did not exist, at least not in a form that reflected the reality of an elite female athlete. So she began building them herself.

Along the way, some of the beliefs she had trained around for years began to unravel.

One example stood out. Faulkner had always assumed that lifting too close to a hard bike session would compromise her performance. In practice, she found the opposite.

“I often rode better when I did gym work first, because it activated muscle groups I wasn’t recruiting enough on the bike,” she said, speaking to Domestique. The effect held even when sessions were moved around, suggesting it was not simply a matter of fresher legs.

It changed how she thinks about activation, and more broadly, how individual those responses can be.

That process exposed a wider issue. Much of performance science, Faulkner argues, still treats female physiology as if it follows a single, predictable pattern.

“People still talk about ‘the female cycle’ as if it’s universal,” she said. “It’s not. Symptom profile, energy availability, fluid retention, even perceived exertion can vary widely between athletes and over time.”

Historically, most performance research has been conducted on men, leaving female athletes to adapt insights that were never designed for them. For Faulkner, that gap became the starting point.

Her background made the leap less improbable than it sounds. A computer science graduate from Harvard University with experience in venture capital and active investments in AI, Faulkner has long operated across disciplines.

She began with what she had: nine years of biometric data. Heart rate, HRV, sleep, power, temperature, training load, menstrual cycle phases, bloodwork, DEXA scans. Each tool offered a fragment. The value, she believed, lay in the connections.

The system she built draws from those inputs and runs against more than 4,000 hours of training history, generating evolving models of her physiology. Not dashboards, but decisions.

Her approach remains deliberately grounded. “Sensations first, models second,” she said. “If my model says one thing and my body says another, I trust my body and use that to improve the model.”

The early returns are difficult to ignore. Faulkner used the system in the lead up to the Pan American Championships, where she won three gold medals. More recently, she produced her best ever twenty minute power.

She is careful not to overstate causation. In real world training, variables rarely move in isolation. Weather, fatigue, cycle phase, nutrition and stress all shift at once.

“In science, you want to isolate one variable,” she said. “In reality, everything changes together. The models improve as the data deepens.”

The idea of athlete specific modelling is still emerging, but Faulkner believes the direction is clear.

“The best teams are moving that way,” she said, pointing to her current team, EF Education-Oatly, where athlete health, including menstrual health and nutrition, is actively prioritised.

Better individual data, she argues, not only sharpens performance but also helps athletes advocate for themselves and identify risks earlier.

For all its promise, the system still runs into limitations. Chief among them is the lack of consistent, structured menstrual data at scale.

Some athletes track diligently, many do not. Access across platforms is inconsistent, and while certain wearables attempt to infer cycle signals, adoption remains uneven.

“As more women choose to share that context alongside performance data, the insights will become much stronger,” she said.

There is also the question of balance. Can an athlete become too data driven in a sport that still depends on instinct?

Faulkner does not see it that way. The risk, she suggests, only emerges when data overrides the athlete rather than supporting them. Data should help explain what the body is feeling, not dictate it.

For now, the system remains a personal tool. But Faulkner is open to taking it further, carefully.

“I’d like to make it available more broadly, especially for women,” she said. “But it has to be done responsibly. Health data is deeply sensitive.”

With the Los Angeles 2028 Olympics on the horizon, Faulkner is effectively running a live experiment on herself. In a sport where margins define outcomes, her competitive edge may be coming from a place the rest of the peloton has barely begun to explore.