Antoine Plissonneau

Design, development, and deployment of Reinforcement Learning solutions for two use cases: smart grid/microgrid control and intraday trading (energy market). Implication from scoping to deployment.

• 𝗥𝗟 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗼𝗿𝘀 & 𝗲𝗻𝘃𝗶𝗿𝗼𝗻𝗺𝗲𝗻𝘁𝘀: microgrid and market engines, modular architecture, testing, packaging, 𝗖𝗜/𝗖𝗗. Created a 𝗿𝗲𝘂𝘀𝗮𝗯𝗹𝗲 𝘁𝗼𝗼𝗹𝗯𝗼𝘅 for energy projects (microgrids & trading).

• 𝗘𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁𝗮𝗹 𝗽𝗿𝗼𝘁𝗼𝗰𝗼𝗹 𝗱𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: model training, tuning, and validation.

• 𝗠𝗟𝗢𝗽𝘀: deployment and monitoring of agents on 𝗧𝗼𝘁𝗮𝗹𝗘𝗻𝗲𝗿𝗴𝗶𝗲𝘀-𝗼𝗽𝗲𝗿𝗮𝘁𝗲𝗱 𝘀𝗶𝘁𝗲𝘀.

• 𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗹𝗲𝗮𝗱𝗲𝗿𝘀𝗵𝗶𝗽: supervision of contractors and interns.

• 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻: internal presentations, documentation, and a publication at the Reinforcement Learning Conference (𝗥𝗟𝗖) workshop.

• 𝗗𝗼𝗺𝗮𝗶𝗻𝘀: model-free RL, Transformers, generalization, sim-to-real, Safe RL.

• Development of time-series 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹𝘀 (ML/RNN) for vehicle pollutant emissions.

• Strategies for data-collection optimization and domain transfer/adaptation of Deep Learning models (𝗔𝗰𝘁𝗶𝘃𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴, 𝗗𝗮𝘁𝗮 𝗣𝗿𝘂𝗻𝗶𝗻𝗴, 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴, 𝗠𝗲𝘁𝗮-𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴).

• 𝗞𝗲𝘆𝘄𝗼𝗿𝗱𝘀: Forecasting, Domain Transfer (Domain Adaptation), Few-shot Learning, Docker, MLflow, AWS.

Industrial PhD (CIFRE) candidate on the Autonomous Freight Train project.

• Doctoral research conducted in collaboration with SNCF, Alstom, Capgemini, and Hitachi.

• Additional work in Computer Vision: obstacle detection (YOLO), rail segmentation (U-Net), and distance estimation.