Super excited to share we won the Grand Prize (1st Overall) at HackCMU amongst 200+ teams!
David Chung, Dheeraj Vislawath, Julian Ng‑Thow‑Hing, and I built medicly (www.medicly.dev), an AI‑powered platform for patients and clinicians to track recovery together.
Last week, David dislocated his shoulder for the 8th time. During this hackathon, I worked remotely because of a hip injury (hence the Photoshop). Neither of us has gone to PT due to the inconvenience and scheduling conflicts. The average wait to get an evaluation is 14 days, and only about 35% of people stick with treatment. We decided to fix that.
Here’s medicly in action:
- 📷 Single Camera Capture + CV: Uses smartphone video and pose estimation to analyze exercises in near real‑time.
- ⏰ Time‑series Kinematics: Splices keyframes, tracks joint angles over time, and records biomechanics metrics such as max/min knee and hip flexion.
- ➰ Claude + CV Feedback Loop: Turns recorded data into clear biomechanics findings, such as poor joint health, unstable flexion, asymmetry, and range‑of‑motion metrics.
- 🏥 Clinicians Keep Control: 3D visuals and exercise recommendations are sent to clinicians to approve/edit. Patients see the next steps only after sign‑off.
Special thanks to the HackCMU organizers and sponsors Citadel, Hudson River Trading, Jane Street, Anthropic, Y Combinator, and more for an awesome event!
Check out our demo here: https://www.medicly.dev/
Source code: https://github.com/scrappydevs/medicly