MOHAMMED IDRI

– Designed and trained deep learning models (CNNs, VAES, Vision Transformers) in PyTorch to predict mechanical properties of industrial materials. – Led research on dynamic polymer characterization. Presented results at SKM23 (Mar 2023). – Co-inventor of a patented deep learning method (FR3153417) for mechanical stress prediction in metallic alloys, achieving 4% RMSE(state-of-the-art performance).

– Giving a 21 hours Machine Learning course to college students.

– Conducted applied AI research to develop deep learning models for mitosis detection (YOLOv5) and tumor tissue segmentation using state-of-the-art PyTorch architectures on large-scale histopathology datasets. – Improved model robustness to staining variability and imaging artifacts through data augmentation and domain adaptation techniques. – Implemented federated learning pipelines to ensure client data privacy while enhancing model generalization. – Maintained and optimized the company's ML codebase, improving reliability, scalability, and experiment reproducibility.

– Design and implementation of a decision-making system (Datamart) for the human resources department using SQL Server and Talend. – Design reports and dashboards displaying strategic KPIs for the HR activity on Tableau Software. – Implementation of Resume Information Extraction (OCR) and a skills matching system (NLP).

– Development and implementation of Genetic Algorithms to solve the infamous NP-Hard Vehicle Routing Problem with Pickup and Delivery (VRPPD) using Google's framework Or-tools. – Development of hybrid optimization technique with K-NN algorithm and integer programming.

– Science popularizer: Using experiments and games to bring youngsters and adults closer to science.

– Giving physics lessons at my local orphanage.