V
Loading...
Loading...
Loading...
Protecting Africa's Financial Revolution
Our real-time fraud detection system processes 15M+ daily transactions across East Africa, protecting users and financial institutions from sophisticated fraud schemes.
$50M
Annual Savings
70%
Fraud Reduction
50%
Fewer False Positives
15M+
Daily Transactions
Mobile money has revolutionized financial services in Africa, with over $1 trillion in annual transactions. But this growth has attracted sophisticated fraudsters who exploit vulnerabilities in the system.
Traditional rule-based fraud detection can't keep up with evolving fraud patterns. Working with major mobile money operators, we developed an AI-powered fraud detection system that analyzes transactions in real-time, identifies suspicious patterns, and blocks fraudulent activity before funds are lost.
The system processes over 15 million transactions daily with sub-100ms latency, preventing an estimated $50 million in fraud losses annually for our partners.
Understanding the unique obstacles we're working to overcome.
Transactions must be scored in under 100ms to avoid degrading user experience.
Fraudsters constantly adapt their techniques, requiring models that can detect new patterns.
Blocking legitimate transactions damages user trust and business relationships.
System must handle 15M+ daily transactions with high availability.
The methods and techniques we've developed to address these challenges.
Distributed ML inference system scoring every transaction in under 100ms.
Network analysis to identify suspicious clusters and money laundering patterns.
Analysis of user behavior patterns to detect account takeover.
Models automatically retrain on new fraud patterns identified by investigators.
Measurable outcomes from our research and deployments.
$50M
Annual Savings
Partner institutions save over $50 million annually in prevented fraud losses.
70%
Fraud Reduction
Overall fraud losses reduced by 70% after system deployment.
50%
Fewer False Positives
ML models reduce false positive rate by 50% compared to rule-based systems.
15M+
Daily Transactions
System processes over 15 million transactions daily across multiple operators.
Kimani, D., Mwangi, J., et al.
Federated learning system sharing fraud intelligence across operators without sharing raw data.
Partners:
M-Pesa, Airtel Money, MTN Mobile Money
James Mwangi
Technical Lead
David Kimani
ML Engineering Lead
We're always looking for collaborators, partners, and talented researchers to advance this work.