SevaSetu (सेवासेतु)

Seva means care. Setu means bridge. SevaSetu is a care bridge between patients, nurses, and doctors in busy clinics. It works in Hindi and English.

The Problem

• Doctors spend 3 to 4 minutes of a 5 minute visit asking the same basic questions.
• After the visit, patients often do not understand their prescription. They forget doses or mix up old and new medicines.

What It Does

Before the Visit

• The patient or nurse speaks or types symptoms in Hindi, English, or Hinglish.
• The nurse can add vitals like BP, temperature, SpO2, and pulse.
• AI asks 3 to 4 simple follow-up questions, like when it started and how bad it is.
• The doctor gets a short brief they can read in about 15 seconds, with a triage tag: Routine, Priority, or Urgent.

After the Visit

• The patient takes a photo of the prescription. No typing needed.
• AI reads the medicine names, dose, timing, and number of days.
• Each medicine gets a simple card in Hindi and English that explains what it is for.
• One tap adds reminders to Google Calendar. Old medicine reminders are removed automatically.
• A .ics file download works even without a Google login.

It is not an AI doctor. It does not diagnose or prescribe. A human checks every AI output.

Value It Adds

For one doctor seeing about 960 patients a month (estimates, to be checked in a real pilot):

• Doctor time saved: 48 to 64 hours per month.
• More patients served: 290 to 385 extra patients per month, with no extra staff.
• Family effort saved: about 300 hours per month of setting alarms and searching drug names.
• Earlier care: urgent cases are flagged and seen sooner.
• Better adherence: fewer missed or double doses, so fewer repeat visits.

The saved doctor time can either end the clinic day about 2 hours earlier or serve more patients. Not both fully.

Tech Stack

• Frontend and backend: Next.js (App Router), TypeScript, Tailwind CSS
• AI: Gemma 4 for text and prescription image reading
• AI hosting: DigitalOcean Gradient, Google AI Studio, or local Ollama, with an offline fallback
• Database: SQLite (better-sqlite3)
• Voice input: Web Speech API
• Calendar: Google Calendar API and .ics export
• Deploy: Docker on DigitalOcean App Platform