💊 Introduction: The Silent Epidemic Killing Millions 🧠 The Vision: AI as a Public Health Guardian ⚙️ How I Built It: A Multi-AI Hybrid System Powered by Amazon Nova Ai 🔹 Core Architecture 🔹 The AI Brain: Multi-Provider Fallback System 🔍 Intelligent Drug Verification System
- 📷 Smart Scanning with OCR + Vision AI (Amazon Nova AI)
- 🧬 Semantic Matching with Transformers
- 🔄 Real-Time Regulatory Sync
⚡ Why Amazon Nova Matters in This Project 🚧 Challenges & Solutions ❗ OCR Limitations on Medicine Packaging ❗ Serverless AI Scaling ❗ Critical System Failures (JSON.parse Error) 📚 What I Learned
- 🛡️ Resilience is Everything
- ⚖️ Speed vs Accuracy Tradeoff
- 🤝 Trust is a Feature
🌍 Real-World Impact: Saving Lives at Scale 🚀 Driving Adoption in Nigeria & Beyond 🛠️ Tech Stack 🔗 Try the App 💡 Final Thoughts
💊 Introduction: The Silent Epidemic Killing Millions
💊 Introduction: The Silent Epidemic Killing Millions
Counterfeit pharmaceuticals are not just a market problem—they are a public health emergency . Across sub-Saharan Africa, over 500,000 deaths annually are linked to fake drugs. In Nigeria alone:
267,000 deaths yearly from substandard malaria medication
169,000 child deaths from fake antibiotics used for pneumonia
A crisis so severe that regulators report it rivals casualties from war
This is not just data—it’s real lives lost.
That’s why I built the Fake Detector App —an AI-powered mobile solution designed to empower everyday Nigerians to verify medications instantly .
🧠 The Vision: AI as a Public Health Guardian
🧠 The Vision: AI as a Public Health Guardian
The goal is simple but powerful:
Put a drug verification system in every citizen’s pocket
By combining cutting-edge AI with real-time regulatory data, this project aims to:
Detect counterfeit drugs instantly
Reduce preventable deaths
Build trust in healthcare systems
Enable informed decisions at the point of purchase
⚙️ How I Built It: A Multi-AI Hybrid System Powered by Amazon Nova Ai
⚙️ How I Built It: A Multi-AI Hybrid System Powered by Amazon Nova Ai
This project leverages a resilient, multi-provider AI architecture , with Amazon Bedrock and Amazon Nova playing a critical role.
🔹 Core Architecture
🔹 Core Architecture
Frontend & Mobile: Next.js 14 + Capacitor (Android-ready)
Backend: Node.js + Prisma + PostgreSQL
Deployment: Vercel + Firebase integrations
🔹 The AI Brain: Multi-Provider Fallback System
🔹 The AI Brain: Multi-Provider Fallback System
To ensure reliability, I designed a custom AI Fallback Manager that orchestrates:
Google Gemini 2.5 Flash
Anthropic Claude 3.5
Amazon Nova (via Bedrock)
👉 If one AI fails (quota, latency, OCR issue), another automatically takes over .
This ensures:
⚡ Faster response times
🛡️ Zero single-point failure
🔁 Continuous availability in critical scenarios
🔍 Intelligent Drug Verification System
🔍 Intelligent Drug Verification System
1. 📷 Smart Scanning with OCR + Vision AI (Amazon Nova AI)
- 📷 Smart Scanning with OCR + Vision AI (Amazon Nova AI)
Users scan drug packaging using their phone camera. The system:
Enhances image quality (lighting, sharpness)
Extracts batch numbers, manufacturer info
Uses AI vision models for better accuracy on curved/metallic surfaces
2. 🧬 Semantic Matching with Transformers
- 🧬 Semantic Matching with Transformers
Using Transformers.js (all-MiniLM-L6-v2) :
Extracted text is compared against thousands of official drug alerts
Detects matches even with:
Misspellings
Formatting inconsistencies
Partial data
3. 🔄 Real-Time Regulatory Sync
- 🔄 Real-Time Regulatory Sync
A custom engine continuously monitors:
NAFDAC recall lists
Public safety alerts
This ensures the app always has an up-to-date “Watchlist” of dangerous drugs .
⚡ Why Amazon Nova Matters in This Project
⚡ Why Amazon Nova Matters in This Project
The Amazon Nova models (via Bedrock) provide:
🧠 Advanced reasoning for ambiguous drug labels
🖼️ Vision capabilities for complex packaging
⚡ Low-latency inference for real-time scanning
🔒 Enterprise-grade reliability
Impact: Amazon Nova acts as a critical fallback intelligence layer , ensuring that even in poor network or high-load conditions, users still receive accurate results.
🚧 Challenges & Solutions
🚧 Challenges & Solutions
❗ OCR Limitations on Medicine Packaging
❗ OCR Limitations on Medicine Packaging
Problem: Curved bottles + reflective surfaces
Solution: Built a custom image preprocessing pipeline
❗ Serverless AI Scaling
❗ Serverless AI Scaling
Problem: Heavy ML models in serverless environments
Solution: Optimized ONNX runtime + Webpack configs
❗ Critical System Failures (JSON.parse Error)
❗ Critical System Failures (JSON.parse Error)
Problem: Tesseract crashes breaking user flow
Solution:
Bypassed local OCR
Prioritized cloud-based AI (Amazon Nova, Gemini, Claude)
📚 What I Learned
📚 What I Learned
1. 🛡️ Resilience is Everything
- 🛡️ Resilience is Everything
A health-tech system must never fail silently . Always have:
Backup AI
Fallback logic
Redundant pipelines
2. ⚖️ Speed vs Accuracy Tradeoff
- ⚖️ Speed vs Accuracy Tradeoff
In life-critical apps:
Accuracy must NEVER be sacrificed for speed.
3. 🤝 Trust is a Feature
- 🤝 Trust is a Feature
Users must feel:
Safe
Confident
Informed
Clear UI/UX and transparent results are essential.
🌍 Real-World Impact: Saving Lives at Scale
🌍 Real-World Impact: Saving Lives at Scale
This project has the potential to:
Reduce counterfeit drug circulation
Save hundreds of thousands of lives annually
Support healthcare workers and pharmacists
Strengthen regulatory enforcement
🚀 Driving Adoption in Nigeria & Beyond
🚀 Driving Adoption in Nigeria & Beyond
To ensure real-world usage:
📱 Mobile-first design (Android focus)
🌐 Lightweight + offline-tolerant architecture
🤝 Partnerships with pharmacies & NGOs
📢 Awareness campaigns on drug safety
Future plans include:
SMS-based verification for low-end devices
Integration with national health systems
Expansion across Africa
🛠️ Tech Stack
🛠️ Tech Stack
Next.js, React, TypeScript
Node.js, Prisma, PostgreSQL
Transformers.js
Firebase, Vercel
Amazon Bedrock (Nova AI)
Google Gemini, Anthropic Claude
🔗 Try the App
🔗 Try the App
🌐 https://scan.sampidia.com
📱 Google Play Store
💻 GitHub Repository
💡 Final Thoughts
💡 Final Thoughts
The Fake Detector App proves that AI is not just about innovation—it’s about impact .
By leveraging Amazon Nova and a resilient multi-AI architecture , this project transforms smartphones into life-saving tools .
In a world where fake drugs can kill, verification should be instant, accessible, and reliable.
