About the challenge

Care is bottlenecked on attention, not knowledge: the radiologist who misses the forty-first scan, the clinic with no doctor, the chronic patient nobody checks on.

AI for Medicine: A Global Student Hackathon on the Future of Care is about pointing AI at that bottleneck instead of adding to it. Build something that helps people notice, measure, or act on a health problem. Show where it works and where it fails.

Hosted online by Open Health AI Collective. All event times are Eastern.

The one requirement

It must be evaluated. A demo is not evidence.

State what you measured, how you measured it, and what the result means. Compare your output with a reference, a baseline, or a stated cost. Name the conditions under which your result stops being useful.

Missing an Evaluation section does not disqualify your entry, but it caps Technical Depth and Rigor & Reproducibility at half marks each.

Who this is for

You must be 18 or older and currently enrolled at an accredited institution. Any level and any field are welcome. Work alone or in a team of up to four.

Every finalist gets a participation certificate. Winners get a track certificate. All certificates are signed PDFs you can cite on a resume. Prizes are certificates only, with cash value $0.

Tracks

  • Best Overall: Highest combined judging score.
  • Best Clinical Relevance: Strongest connection to a real care problem.
  • Best Rigor & Reproducibility: Clearest evaluation and most reproducible work.
  • Best Novel Data Source: Most useful approach to collecting or using a distinctive source of data.
  • Best Demo: Clearest demonstration of a working project.
  • People's Choice: Decided by public vote.

The organizer panel decides the five judged tracks. Public vote totals never affect those tracks.

Get started

  1. Pick one care problem and identify who has it.
  2. Choose data you can use and a reference you can measure against.
  3. Build a working pipeline and evaluate it.
  4. Publish your repository, record your video, and write up your method, results, and limitations.
  5. Submit by Friday, December 4, 2026, 11:45 PM ET.

Requirements

What to Build

Build an AI, data, or signal-processing project that addresses a problem in health or care. Give it a defined task and evaluate how well it performs that task.

New, in-progress, and previously published projects are allowed. Your repository must be public, and your project must run today.

What to Submit

  1. A public repository URL with a README. Explain what the project does and how to run it. Include dependencies, configuration, and instructions needed to reproduce your evaluation.
  2. A video of at least 60 seconds showing it working end to end. Show a working result within the first 60 seconds. A viewer must be able to watch without installing anything.
  3. A written description. Include the problem and who has it; your method; an Evaluation section stating your metric and measurement method; a Who benefits, and who could be harmed section; limitations; and what you would do with another month.
  4. Full attribution. Credit every model, dataset, and piece of code you did not write.

Missing an Evaluation section does not disqualify your entry, but it caps Technical Depth and Rigor & Reproducibility at half marks each.

Hackathon Sponsors

Prizes

6 non-cash prizes
Best Overall
1 winner

A signed PDF track certificate for the highest combined judging score. No cash value.

Best Clinical Relevance
1 winner

A signed PDF track certificate for the strongest connection to a real care problem, with claims supported by evidence. No cash value.

Best Rigor & Reproducibility
1 winner

A signed PDF track certificate for the clearest evaluation and most reproducible work. No cash value.

Best Novel Data Source
1 winner

A signed PDF track certificate for the most useful approach to collecting or using a distinctive source of data. No cash value.

Best Demo
1 winner

A signed PDF track certificate for the clearest demonstration of a working project. No cash value.

People's Choice
1 winner

A signed PDF track certificate for the entry selected by public vote only. No cash value.

Devpost Achievements

Submitting to this hackathon could earn you:

Judges

Nolan Hamilton

Nolan Hamilton
Dr. Hamilton

Judging Criteria

  • Technical Depth — 25%
    Is there a real pipeline behind the interface? Signal processing, fine-tuning, a custom dataset, or a retrieval or inference strategy all count. A prompt and a UI with no evaluation do not.
  • Rigor & Reproducibility — 25%
    Use a held-out evaluation and report a metric in clinical units or a stated cost. Include seeds, configurations, and code. Can a stranger rerun your evaluation and understand any differences?
  • Clinical Relevance & Honesty — 25%
    Address a real problem and scope your project to what it can support. Name the limitations and who could be harmed by an error. Your accuracy claim must match your evidence.
  • Demo Quality — 25%
    Show the project visibly working within the first 60 seconds. Make the input, process, and result understandable. A stranger must be able to follow the demonstration without installing anything.

Questions? Email the hackathon manager

Invite others to compete

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.