BRUNO STORDEUR

Built and deployed scoring and fraud-detection ML systems for an instant-payment and digital-wallet platform where fraud prevention is critical to user trust. Developed real-time fraud scoring models and designed “cold start” rules for first-time users to reduce false positives. Conducted exploratory analysis of suspicious behaviors across transaction types (P2P, e-/m-commerce). Ran production monitoring and performance tracking with IBM Safer Payments, including dynamic threshold tuning. Implemented scientific monitoring and benchmarking to propose robust methods against evolving fraud patterns.

Supported operational risk management through historical incident analysis to target risk areas and reduce costs. Structured an internal incident database by risk categories (e.g., fraud, human error) and linked incidents to root causes including processes, systems, and failing business units. Analyzed correlations to detect recurring patterns across risk perimeters. Produced clear visualizations and reporting for non-technical stakeholders and contributed to business review committees by communicating insights and recommendations.

Worked on ML/DL methods for clinical decision support to help treat intracranial hypertension events in patients. Generated synthetic temporal data using a DGAN model to address rarity and imbalance. Defined sequence quality metrics using Dynamic Time Warping to filter usable synthetic sequences. Modeled arterial and intracranial pressure with step-by-step predictive CNN-LSTM networks and classified hypertension episodes using multiple architectures (TCN, CNN, LSTM, Random Forest, etc.). Maintained ongoing scientific monitoring, testing, and benchmarking of state-of-the-art approaches validated with clinicians.