Inspiration
A private OB-GYN practice in India can follow hundreds of pregnancies on paper registers. One missed anomaly scan or skipped follow-up can mean a missed complication, and nobody notices until it is too late. We wanted a tool that helps a small clinic keep every mother in follow-up, in the languages the clinic actually works in.
What it does
MatruSetu is an assistive pregnancy-tracking and OPD-digitisation app for OB-GYN clinics, available in English, Hindi and Marathi.
- Auto ANC schedule: enter the LMP once and the FOGSI/MoHFW-based schedule of scans, tests and injections, gestational age and EDD build themselves.
- Follow-up risk: every mother is marked on track, at risk or lost, based on overdue items and contact history, with clinical alert filters.
- One-tap follow-up: a call queue with one-tap call or WhatsApp reminders in the mother's language, plus contact logging.
- Visit recording: vitals with clinical flags (BP, Hb, FHR, fundal height) and trend charts, and a one-page printable patient report.
- Digitised paper: photographed case papers, PDFs and lab reports become one searchable timeline through in-browser OCR.
- Voice AI (optional, Gnani): dictate vitals in Hindi, Marathi, English or Hinglish, play spoken counselling to mothers, and handle WhatsApp voice notes with a danger-sign screen that replies urgently and flags staff.
- Works offline: an installable PWA whose visits and call logs queue offline and sync later.
How we built it
Next.js 16 (App Router), React 19, TypeScript and Tailwind CSS v4 on the front end; Supabase (Postgres, Auth, Storage) with row-level security on every table; Serwist for the PWA and offline queue; Tesseract.js and pdfjs-dist for in-browser OCR; Gnani AI Prisma (speech-to-text) and Timbre (text-to-speech) for voice, with browser speech APIs as a fallback.
Challenges we ran into
Making voice dictation reliable across Hindi, Marathi and Hinglish (we wrote a Devanagari-aware parser, and the clinician confirms before anything is saved), building an idempotent offline write queue, and keeping every feature failing closed when configuration is missing.
Accomplishments that we're proud of
590 unit tests across 33 files, a CI pipeline with lint, typecheck, SQL validation and a dependency audit (0 high vulnerabilities), full English/Hindi/Marathi coverage with key-parity tests, and accessibility work (AA contrast, 44 px touch targets, keyboard and screen-reader support).
What we learned
Continuity of care fails at the follow-up step rather than at diagnosis, and the language and connectivity of a real clinic matter as much as the clinical logic.
What's next for MatruSetu
Clinical sign-off of the schedule and risk thresholds by a practising OB-GYN, review of the Hindi and Marathi wording by native-speaker clinicians, persisting the pregnancy questionnaire to the database, and proactive outbound WhatsApp reminders once Meta Business verification is complete. MatruSetu is assistive, not diagnostic.
Team: Madhav Zanwar, Veerbhadra Mahant, Divyani Papalkar, Hari Chaudhari. MatruSetu was first built for the Koita Foundation / IIT Bombay Health-a-thon 2026 (Maternal & Child Health track).
Live demo: https://matruseva.vercel.app (demo clinic with 10 synthetic patients; demo login is pre-filled on the sign-in page) Code: https://github.com/madhavzanwar/Matruseva
Built With
- gnani-ai
- next.js
- playwright
- postgresql
- pwa
- react
- supabase
- tailwind-css
- tesseract.js
- typescript
- vercel
- vitest
