Undiagnosed Patients Hackathon
International rapid-prototyping hackathon bringing AI, clinicians, and patients together to solve undiagnosed rare disease cases in 48 hours.
Michael Brooks and I designed the first Rare Disease AI Hackathon at Stanford. The concept: walk up to physician heroes and ask, “How can I help?” Then build multi-expert teams — clinicians, data scientists, patients, engineers — and tackle real diagnostic cases in 48 hours. Small squads, real problems, real people waiting for answers.
It grew into an international movement. The Karolinska Institute in Stockholm hosted the inaugural Undiagnosed Hackathon in 2023, delivering four diagnoses from ten families. In 2025, Mayo Clinic hosted 130 collaborators from 28 countries. Six diagnoses in 48 hours. Six families who walked in without answers and walked out with a name for what they’d been carrying.
The model works because it inverts the usual power dynamics. Patients aren’t subjects — they’re collaborators. Clinicians aren’t gatekeepers — they’re teammates. And the AI tools serve both, accelerating the pattern-matching that human experts do best but can’t always do fast enough.
Every hackathon teaches us something we didn’t know. That’s the point. The format is designed to learn as much as it is to solve. When we got the support we needed from our AI friends, the results spoke for themselves — and the movement keeps growing.