Real-time movement feedback. Right in the browser.
A sports platform pairing people to train together, with an AI trainer that gives feedback on movement in real time.
01 / The problem
Sportify starts from two problems in physical activity. First, people do not always have a partner to exercise with. Second, when they train alone they do not always know whether their movement is correct.
The platform addresses both at once through partner matchmaking, open play sessions, and an AI trainer that gives feedback on exercise form.
02 / Decisions
Native PHP instead of a framework
The team had uneven levels of technical experience. Requiring everyone to learn a new framework would have spent time we needed for shipping features inside a fixed funding timeline. So we chose native PHP without a framework. The cost is structure we have to maintain ourselves.
Pose estimation runs on the user's device
Inference happens in the browser, on the user's own hardware. We rejected sending workout video to a server for processing: it needs heavier infrastructure, and it means a person's video has to leave their device. Running it client-side reduces what the server has to carry and keeps the video local. The trade-off is that accuracy and performance now depend on the device someone happens to own.
03 / The outcome
- Selected for PKM-KC funding at national level.
- Phase one shipped as a working platform, not a concept or a presentation prototype.
- AI trainer with real-time pose feedback, partner matchmaking, open play sessions, and progress tracking.
- Phase two is underway, focused on the interface, pose estimation accuracy, and deployment.