Inspiration

We’ve watched our grandparents struggle to keep track of their own health: medicines, appointments, food restrictions and more. Many apps solve parts of it, but they’re hard to use and none does it all.

The numbers made it bigger than our own families. India has about 149 million people aged 60+, and the WHO estimates that roughly half of people with chronic illness don’t take medicines as prescribed. Many of these people can’t type in English, can’t read messy handwriting, and can’t decode shorthand like 1-0-1, BD or HS. Their children often live in another city and only find out about a missed BP or sugar tablet after a health scare.

We wanted one very simple app that gives older people control over their own health and independence, and keeps their family in the loop.

What it does

Docsfile is a voice-first medicine companion for older people.

Setup from a photo: snap the prescription and Docsfile builds the medicine schedule. Nobody types. Understands Indian prescriptions: 1-0-1, OD, BD, HS, SOS, before or after food. Safe by design: unclear names or timings are flagged, and a person confirms before any reminder is set (it catches things like “Metfornin”). Spoken reminders in Kannada, Tamil, Hindi and English: “Amma, time for your BP tablet, after food.” Closes the loop: a missed dose or unusual watch reading sends the family a WhatsApp alert the same day. “Help me” voice chat for booking a doctor or joining a video call. Doctor visits: appointment reminders, food restrictions and a one-page health summary.

It also helps younger people with diabetes, thyroid or BP, who use the same photo-to-reminder flow.

How we built it

We built a working prototype with Flutter for the phone app. The pipeline is:

Photo: the prescription is read with on-device OCR (ML Kit). Understand: our own parser turns Indian shorthand into a structured schedule, with a vision LLM as backup for messy handwriting. Confirm: the user or family member reviews anything unclear. Remind and alert: offline reminders and voice run on the device, while Firebase, FCM and Twilio WhatsApp handle sync, notifications and family alerts. Voice: Bhashini or Sarvam provide Indian-language speech, kept swappable so we aren’t locked into one AI provider.

The prototype was built with AI-assisted development (Claude), and all sample prescriptions are invented.

Challenges we ran into Prescriptions aren’t standard. Handwriting, abbreviations and brand-vs-generic names vary wildly, so OCR alone isn’t enough. Safety over speed. A wrong reminder is worse than no reminder, so we added a confirm screen that flags uncertain results instead of guessing. Designing for older users. Large text, few menus and voice instead of typing meant rethinking each screen. Multiple Indian languages for natural-sounding spoken reminders. Keeping family alerts useful without overwhelming them. What we learned Great health tech for older people starts with removing steps, not adding features. In healthcare, a human confirming the AI’s reading is a feature, not a weakness. Local-language voice makes the app far more usable for people who never needed to type before. What’s next Field testing with real older users and families More regional languages Deeper smartwatch integration Doctor and pharmacy connections