Machine Learning

Credit Default Prediction — Crédit Agricole Personal Finance & Mobility

Machine learning model (logistic regression, gradient boosting) estimating customers’ probability of default on a 30-million-row dataset, plus a SAS-to-Snowflake database migration for pipeline automation.

Neural network approximation with uniform vs gradient-aware sampling

Neural Networks for High-Gradient Functions — Research Apprenticeship at I2M

Research work-study at the Institut de Mathématiques de Marseille: neural network theory, TensorFlow experiments, and a gradient-aware sampling method that improves approximation where the target function varies sharply.