Offline-first nutrition decision-support app for CHPS frontline health workers in Northern Ghana — UNICEF AI for Nurturing Care Hackathon
# NurtureLink
An offline-first nutrition tool that helps CHPS Community Health Officers in rural Northern Ghana turn the visit data they already collect into affordable, in-season feeding guidance a caregiver can actually act on, delivered in local language. It supports the health worker's judgement rather than replacing it.
Built for the **UNICEF AI for Nurturing Care Hackathon** (KOICA / MEST StartUp Lab).
Bootcamp: 26–28 August 2026, Tamale.
> **Status — working offline-first MVP.** The recommendation engine, encrypted on-device storage (SQLCipher), outbox sync, the severe-case referral guardrail, and 50+ passing engine tests are implemented and runnable against a seeded pilot district. Voice audio playback, the longitudinal trend chart, and the DHIMS2 export are in progress for the 26–28 August bootcamp.
**Team:** Yakubu Lute (Technical Lead) · Leticia Offeibea (Health & Nutrition Domain)
**Jump to:** Screenshots · Recommendation Engine · AI Architecture · Challenge Mapping
**Implemented:** client registration · visit capture · dietary-diversity scoring · deterministic recommendation engine · seasonal + affordability food selection · referral guardrail · offline SQLite with outbox sync · at-rest encryption · JWT/PIN auth · reference-bundle versioning · Express + Prisma backend · pilot-district seed data
**In progress (bootcamp):** Dagbani audio playback · trend chart on client screen · pull-sync UI · DHIMS2 export serializer · field validation of food/threshold data
---
## Screenshots
| | | |
| --- | --- | --- |
| | | |
| **Sign in / offline mode** | **Prioritised caseload** | **Longitudinal nutrition trend** |
| | | |
| **Seasonal feeding plan** | **Voice note delivery** | **Severe-case referral guardrail** |
| | | |
| **Referral tracker** | **Visit history** | **Client registration** |
| | | |
| **Data sync** | **Offline mode** | **Profile & device info** |
> Screenshots captured in development on iOS; production target is Andr …