Mohamed Khalil Jabri

- Collaborate with business stakeholders to define requirements, identify key challenges in areas such as EV charging and Microgrid optimization, and translate them into Reinforcement Learning (RL) use cases with practical, real-world impact. - Develop RL-based intraday trading strategies on European energy markets, optimizing EV-charging by exploiting price fluctuations under imperfect forecasts and complementing day-ahead positions. - Improve existing RL controllers for microgrid operations by extending algorithms to coordinate EV charging dispatch and battery storage management, aiming to reduce operating costs and CO2 emissions.

- Conducted research at the intersection of Reinforcement Learning, Imitation Learning, and Causality. - Developed causality-inspired methods to improve sample efficiency and generalization in imitation learning. - Collaborated with industry and academic partners on applied machine learning projects, and actively promoted research through presentations, public talks, and interdisciplinary events.

- Developed an explainable vessel recognition system to enhance interpretability in maritime security systems. - Implemented a Vision Transformer model with pairwise heatmap visualization using Attention Flow Approximation. This approach provided intuitive interpretation of results, matching the performance of standard CNN models while offering superior explanations compared to classic techniques like GradCam.

Developed a Reinforcement Learning protocol for network congestion control, achieving performance comparable to conventional heuristics in a simple simulated network.

- Joined the startup in its early stages, contributing to the development of its first POC for a Reinforcement Learning-based decision-making system for logistics. - Contributed as part of a team of three to the development of a Reinforcement Learning-based solution to the Bin Packing problem inspired from DeepMind's AlphaGo, significantly outperforming existing heuristics and solvers. Was mainly responsible for the parallelization and integration of Monte Carlo Tree Search within the AlphaGo framework.

Developed a brand logo recognition system using transfer Learning.