Sofiane Hibbah Falaki

Real-time SEPA payment anomaly detection; developed end-to-end data and ML solutions from ingestion to feature engineering to model deployment on scalable cloud and containerized environments. Implemented automated monitoring workflows via Airflow and container orchestration.

Laid the foundation for a data ingestion and ML inference pipeline for anomaly detection in SEPA payments. Implemented Isolation Forest and clustering for unsupervised anomaly detection, built a FastAPI microservice for ML inference, and containerized with Docker with orchestration via Kubernetes. Established Airflow-driven data workflows, enabling end-to-end data processing (ingestion → ML inference → monitoring). Delivered a production-ready data flow from data to ML to monitoring, including model evaluation (F1 score ~0.84 on labeled data) and performance optimizations.

Applied Mathematics for Data Design: the entire data pipeline including ingestion workflows, data structuring, and behavioral feature engineering for SEPA transactions. Built an unsupervised detection system (Isolation Forest + Clustering) achieving 0.84 F1-score with unlabeled data.