Exploring whether on-device AI can solve real-world medication management friction
Anyone managing multiple medications knows the friction: deciphering GP shorthand, tracking different schedules, remembering what's running low. Tools exist, but the manual entry alone is enough to put most people off. I wanted to see if AI could simply remove that friction. As a solo learning experiment in React Native and on-device AI, I built Dosi to explore whether AI could make medication management genuinely effortless.
The idea
Anyone who's managed multiple medications knows the friction: deciphering GP shorthand, tracking different schedules, remembering what's running low. Tools exist, but the manual entry alone is enough to put most people off.
I wanted to see if AI could simply remove that friction.
How it works
Dosi uses ML-powered OCR and an LLM to extract medication details directly from a photo of the label. No manual entry required. From there, it tracks supply levels, predicts when you'll run out, and sends smart reminders based on your actual schedule.
Key features:
- AI-powered label scanning that pulls medicine name, dosage, and instructions from a photo
- Supply level monitoring with visual indicators and run-out predictions
- Smart dose reminders personalised to your medication schedule
- Complete medication overview showing next doses, frequencies, and supply status at a glance
Where it landed
The app is live on TestFlight. This was a learning project rather than a product launch, so I'm not putting it on the app stores.
That said, I am open to investments from that AI bubble I've been hearing about. Will accept no less than £250m.