MEDi
AI-driven medication detection platform
MEDi is an AI-driven platform that helps visually impaired shoppers identify pharmaceutical products in stores, so that medication is not taken by mistake.
My team and I launched it in February 2025.
Approach
Rather than broad assistance, MEDi gives a step-by-step shopping guide built on hand-tracking and object detection, so the guidance follows where the user is actually reaching. Audible recognition carries the same information without requiring the screen.
Stack
| Area | Tools |
|---|---|
| App | Kotlin, Android |
| On-device | TensorFlow Lite, MediaPipe |
| Optimizing | Model quantization for edge devices |
Running inference on the device rather than in the cloud is what keeps the guidance responsive enough to follow a hand in real time, and quantization is what makes the models small enough to do it.
Interface
Status
MEDi is in beta testing. You can join us by clicking the logo below.
If you’re interested in our code, check out MEDi’s GitHub repository! 🚀