African ID Verification | Computer Vision Over ML Hype

The Human Story: “When someone in Lagos wants to send money to family in Nairobi, they shouldn’t pay $50 in verification fees. I made it cost pennies.”

The Challenge: Chipper Cash (African fintech unicorn) was hemorrhaging money on third-party ID verification for 5 million users across multiple African countries. They needed in-house verification that worked offline—because reliable internet is luxury in many communities they serve.

The Contrarian Approach: While everyone screamed “MORE ML!”, I advocated for robust Computer Vision + OCR. Why? Because when someone’s remittance depends on your system, you need predictable reliability, not black-box magic.

The Solution:

• Offline-First Design: Works without stable internet connection
• Cost Elimination: Removed all third-party dependencies
• Multi-Country Support: Handles ID formats from 2 African nations
• Fraud Detection Bonus: Uncovered $500K in cyclic referral fraud through feature engineering

The Impact:

• Business: Massive cost reduction (figures confidential, but significant)
• Human: Cheaper international transfers for families across Africa
• Technical: 99.2% accuracy rate processing 12,000 IDs/month

Tech Stack: Computer Vision, OCR, Snowflake, DBT, Feature Engineering “Sometimes the best ML solution is no ML at all.”