Overview
AI for Medicine is an online, global student hackathon hosted by the Open Health AI Collective. It asks participants to build AI, data, or signal-processing projects that help people notice, measure, or act on a health or care problem. The challenge starts from the idea that care is limited by attention rather than knowledge, for example when a scan goes unreviewed, a clinic has no doctor available, or a chronic patient is not followed up. Projects can address diagnosis, triage, monitoring, or access.
The event is open to students aged 14 and older who are currently enrolled at an accredited institution, in any field and at any level. Participants may work alone or in teams of up to four. Companies and professional organizations are excluded. Entry is free, and the challenge is open to participants from all countries and territories, subject to standard exceptions.
Key dates
- Start: November 4, 2026
- Submission deadline: Friday, December 4, 2026, 11:45 PM Eastern Time (Dec 5, 2026, 4:45 AM UTC)
All event times are listed in Eastern Time.
What to build
Each project needs a defined task and a working pipeline. Participants should pick a real care problem, identify who has it, choose data they can use, and select a reference or baseline to measure against. New, in-progress, and previously published projects are all allowed, provided the repository is public and the project runs at the time of submission.
Evaluation requirement
The one firm requirement is that the project must be evaluated. A demo alone is not accepted as evidence. Entries should state what was measured, how it was measured, and what the result means. Results should be compared with a reference, a baseline, or a stated cost, and the conditions under which the result stops being useful should be named. An entry without an Evaluation section is not disqualified, but its scores for Technical Depth and Rigor and Reproducibility are capped at half marks each.
What to submit
- A public repository URL with a README that explains the project, how to run it, dependencies, configuration, and how to reproduce the evaluation.
- A written description covering the problem and who has it, the method, an Evaluation section with the metric and measurement approach, a section on who benefits and who could be harmed, limitations, and what the team would do with another month.
- Full attribution for every model, dataset, and piece of code not written by the team.
- A video recording of the project.
Judging criteria
Each of the four criteria is weighted equally at 25%:
- Technical Depth: whether a real pipeline sits behind the interface, such as signal processing, fine-tuning, a custom dataset, or a retrieval or inference strategy.
- Rigor and Reproducibility: held-out evaluation, metrics in clinical units or stated cost, seeds, configurations, and code that a stranger can rerun.
- Clinical Relevance and Honesty: a real problem, a scope matched to what the project can support, named limitations, and an accuracy claim that matches the evidence.
- Demo Quality: a visibly working project within the first 60 seconds, with input, process, and result that are understandable without installing anything.
Awards
There are six non-cash awards, each with one winner, given as signed PDF certificates with no cash value:
- Grand Prize: highest combined score across the four criteria.
- Excellence in Research and Reproducibility: strong evaluation, appropriate comparisons, and transparent limitations.
- Most Promising Venture: a clear customer need, evidence of demand, and a realistic path to adoption. This award does not promise investment or funding.
- Best Logo: originality, clarity, and relevance to the project.
- Best User Interface: clarity, navigation, and usability.
- Best Presentation: the clearest explanation of the project, its working features, and its results.
Finalists receive a participation certificate. The organizer panel selects all award winners, and public voting does not determine them.
Judges
The judging panel includes Dr. Nolan Hamilton.
Questions
Questions can be directed to the hackathon manager by email.




