Sentinelle AI - Real-time fraud detection system for Mobile Money

Built during the Clapay Hackathon, Sentinelle AI is a machine learning-based fraud detection engine designed to secure Mobile Money transactions.

The project covers the full data science pipeline: generating a synthetic dataset of 8,000 transactions via a behavioural scoring system (incoming calls, typing speed, entry point, transaction type), training a Random Forest classifier, and evaluating performance with an F1-score of 87% and an AUC-ROC of 0.96.

The model is exported as a .pkl file and ready to be integrated into a real-time scoring API.

The project can be adapted to a company's real data.