mamaCare bulter

Inspiration

As a new father and healthcare software developer in Kenya, I watched my wife navigate pregnancy with confusing ultrasound reports, forgotten medication schedules, and anxiety between monthly clinic visits. Having deployed OpenMRS systems serving 100,000+ patients and mChanjo immunization tracking across 50+ Kenyan counties, I've seen how 342 mothers die per 100,000 births in Kenya - yet 70% of maternal complications are preventable with early detection. I knew AI and mobile technology could bridge this gap.

What it does

MamaCare Butler is an AI-powered maternal health companion built with Google Gemini 3 that provides:

Weekly AI Health Check-ins: Gemini 3 Flash analyzes symptoms and vital signs to detect pre-eclampsia, gestational diabetes, and other high-risk conditions with personalized risk assessments
Ultrasound Translator: Gemini 3 Flash Vision extracts measurements (BPD, FL, AC, EFW) from ultrasound images and explains findings in simple, mother-friendly language
Smart Kick Counter: Gemini 3 Flash-powered fetal movement tracking with statistical pattern analysis that alerts mothers to potential distress
Medication Tracker: Comprehensive medication management with scheduled reminders for prenatal vitamins, iron tablets, and prescriptions
Emergency SOS: One-tap SMS alerts with GPS location sent to emergency contacts and local hospitals
Health Trends Dashboard: Visual charts tracking blood pressure, weight, and fetal movements with period filtering (1 week to full pregnancy)
Personalized Profile: Complete pregnancy tracking with due date countdown, week-by-week progress, and emergency contact management

How we built it

Frontend:

Flutter 3.5+ for cross-platform mobile (Android/iOS)
Material 3 design with responsive layouts using flutter_screenutil
Custom chart visualizations with fl_chart library
Real-time notifications using flutter_local_notifications

Backend:

Serverpod 3.1.1 with full REST API architecture
PostgreSQL database for secure health data storage

AI Integration (Google Gemini 3):

Google Gemini 3 Flash Preview (gemini-3-flash-preview) for health risk assessment and natural language health advice generation
Gemini 3 Flash Vision for multimodal ultrasound image analysis — extracting measurements like BPD, HC, AC, FL, and EFW from ultrasound scans and providing mother-friendly explanations
Gemini 3 Flash for AI-powered kick count pattern recognition and anomaly detection with clinical guideline-based assessment
Integrated via the google_generative_ai Dart SDK with structured JSON prompting for consistent, medically-relevant responses

Challenges we ran into

AI Medical Safety: Balancing Gemini 3's powerful reasoning with medical disclaimers — ensuring AI provides helpful insights without replacing doctors
Structured AI Output: Engineering prompts for Gemini 3 Flash to return consistent JSON responses for ultrasound analysis while maintaining clinical accuracy
Scheduled Notifications: Android emulator limitations with exact alarm scheduling required creative workarounds and real device testing
Database Migrations: Managing complex health data schema with Serverpod migrations and PostgreSQL compatibility
Real-time Chart Updates: Building responsive health trend visualizations with period filtering and empty state handling
Emergency Services Integration: Implementing reliable SMS and GPS functionality across different Android versions

Accomplishments that we're proud of

Fully Functional MVP: 10+ major features working end-to-end with real Gemini 3 AI integration
Gemini 3 AI Excellence: Successfully integrated Gemini 3 Flash for both text analysis and multimodal vision capabilities — health assessments, ultrasound analysis, and kick pattern detection all powered by Gemini 3
Complete Health Ecosystem: Built comprehensive maternal health tracking — from daily medications to emergency SOS
Production-Ready Backend: Robust Serverpod architecture with 6 REST endpoints, proper error handling, and database migrations
Professional UI/UX: Clean, intuitive interface with responsive design, loading states, and smooth navigation
Kenya-Specific Optimization: Timezone handling, local emergency numbers, and SMS integration for African context
Real Healthcare Experience: Leveraged 5+ years of health tech experience deploying systems serving 100,000+ patients

What we learned

Gemini 3 Flash excels at medical text analysis, multimodal image understanding, and structured JSON output for healthcare applications
WhatsApp is crucial for Kenya (90% penetration vs 30% app usage)
Offline-first is non-negotiable for healthcare apps in Africa
Simple, intuitive UI matters more than fancy features
FHIR compliance enables seamless integration with existing systems