V
Loading...
Loading...
Loading...
AI for Financial Access
Machine learning models for credit scoring, fraud detection, and financial services designed to serve the unbanked and underbanked populations across Africa.
15M+
Users Served
3x
More Approvals
70%
Fraud Reduction
$50M
Annual Savings
Over 60% of adults in Sub-Saharan Africa lack access to formal financial services. Traditional approaches to banking—requiring credit history, collateral, and physical branches—don't work for this population.
AI offers a path forward. By analyzing alternative data sources like mobile phone usage, transaction patterns, and social connections, we can build credit scoring models that work for people without traditional financial history. Our fraud detection systems protect users and institutions while maintaining the speed and convenience that makes mobile money transformative.
Our Financial Inclusion AI research combines cutting-edge machine learning with deep understanding of African financial systems and user needs.
Understanding the unique obstacles we're working to overcome.
Traditional credit scoring fails for the 60%+ of African adults without formal banking history.
Mobile money fraud costs the industry hundreds of millions annually, threatening trust in digital finance.
Even alternative data sources may be sparse for new users or those with irregular income patterns.
ML models risk encoding biases that could unfairly exclude vulnerable populations from financial services.
The methods and techniques we've developed to address these challenges.
Credit models using mobile money transactions, phone usage patterns, and other alternative data sources.
Graph neural networks and behavioral analysis for detecting fraud in sub-100ms latency.
Techniques to ensure models don't discriminate against protected groups while maintaining predictive power.
Models that can explain their decisions to users, regulators, and compliance teams.
Measurable outcomes from our research and deployments.
15M+
Users Served
Our credit scoring models have enabled loans for over 15 million previously unbanked users.
3x
More Approvals
Alternative credit scoring enables 3x more loan approvals compared to traditional methods.
70%
Fraud Reduction
Our fraud detection systems reduce losses by 70% for partner institutions.
$50M
Annual Savings
Partner institutions save over $50M annually through our fraud prevention technology.
Mwangi, J., Okonkwo, A., et al.
Kimani, D., Mwangi, J., et al.
ML platform for credit scoring using mobile money and telco data, deployed with multiple lenders.
Partners:
Branch International, Tala
Fraud detection system processing 15M+ daily transactions across East African mobile money networks.
Partners:
M-Pesa, Airtel Money
James Mwangi
Research Lead
David Kimani
ML Engineering Lead
We're always looking for collaborators, partners, and talented researchers to advance this work.