VeriMed

Inspiration

The inspiration for VeriMed came from witnessing people selling and buying fake medications in Africa due to lack of resources and regulations. Abdoul has seen expired medicines being sold in villages, and a shocking statistic stands out: over 250,000 people die annually from counterfeit drugs, with up to 70% of medications being fake in some developing countries. As a team passionate about using technology to solve real-world problems, we were deeply moved by the thought that people are literally dying from fake medicines while we have the technology to prevent it. The idea of using smartphone cameras and AI to instantly detect counterfeit medicines felt like a perfect solution accessible, immediate, and life-saving.

What it does

Instant Detection: Point your phone at any medicine and get real-time authenticity verification AI Analysis: Uses multiple AI models to analyze packaging, pill appearance, and batch codes Offline Functionality: Works anywhere without internet connectivity Solves Critical Problem: 70% of medications are fake in some developing countries Global Accessibility: Works on any smartphone worldwide Life-Saving Potential: Prevents consumption of dangerous counterfeit medicines

How we built it

Our team approached this as a systematic, phase-based development process:

Phase 1: Mobile Foundation

Set up React Native/Expo project with TypeScript
Implemented camera integration for medicine scanning
Built user authentication and role-based access
Created professional UI/UX with smooth navigation

Phase 2: Data Collection & ML Setup

Developed synthetic data generation system
Built real-world data collection pipeline
Implemented image preprocessing and augmentation
Created quality assessment and metadata tracking

Phase 3: ML Integration & Testing

Designed multi-model CNN architecture (packaging, pill, batch code, fusion)
Implemented real-time inference service with on-device processing
Built comprehensive training pipeline with progress tracking
Created model testing suite with performance validation
Developed professional ML dashboard for training and monitoring

Challenges we ran into

Technical Challenges Dependency Hell: React Native/Expo projects have complex dependency requirements. We spent significant time resolving peer dependency conflicts, especially with TensorFlow.js and camera libraries.

Mobile ML Performance: Getting ML models to run efficiently on mobile devices was challenging.

What we learned

This project was an incredible learning journey across multiple domains:

Technical Skills
React Native & Expo: Built a complete mobile application
Machine Learning: Implemented real-time ML inference using TensorFlow.js and TensorFlow Lite
Computer Vision: Developed custom CNN architectures for packaging, pill, and batch code analysis
TypeScript: Achieved 97.3% type coverage for robust, maintainable code
Firebase Integration: Set up authentication, database, and storage services

ML & AI Concepts

Multi-model Architecture: Designed separate CNNs for different analysis types
Fusion Networks: Combined multiple model outputs for final decisions
User Experience: Created intuitive interfaces for complex ML operations ## What's next for VeriMed