Mobile · Edge AI · Newborn Care
LumiCare
Offline-first newborn care support for noticing concerns, keeping records, and following up.

01 · Overview
LumiCare is a mobile prototype for caregivers and community health workers. It brings newborn jaundice education, a guided eye-photo workflow, on-device records, care contacts, and follow-up reminders together in one app. Its aim is to make the next step easier to understand and keep important care information available when connectivity is limited. Built with Favour Oni for the MTN Pachipanda Challenge — and won. I developed the mobile app.
02 · The problem
Caregivers and community health workers often work where connectivity is limited, yet newborn concerns like jaundice need clear guidance, good records, and timely follow-up with a qualified professional.
03 · What I built
My part: the whole mobile app. An Expo / React Native app where every scan runs a bundled ONNX model on-device: a YOLOv8 segmentation model finds the sclera, a yellow-pixel index is computed from the mask, and a ported linear classifier maps it to an evaluation-only display state. Records live in an encrypted on-device SQLite database, with PDF exports, referral sharing, and follow-up reminders.
Key features
- Guided eye-photo workflow — camera or gallery
- On-device ONNX Runtime inference (YOLOv8 sclera segmentation)
- Offline-first records with SQLCipher-configured SQLite
- PDF export and referral sharing
- Follow-up reminders with native notifications
- Newborn jaundice education and danger-sign guidance
- Language selector with 20 options, including Yoruba, Igbo and Hausa
Tech stack
Origins
LumiCare is an engineering-evaluation prototype — not a medical device, and its model is not clinically validated.


