2022 - Now
Research Scientist & Consultant at Talan
Paris, France
Lead of several R&D projects. Design of Graph Machine Learning-based algorithms for the generation of synthetic tabular data. Leveraged state-of-the-art variational autoencoders and adapted loss functions.
Consultant at Enedis. Contributed to the development of a novel machine learning approach to estimate the missing amount of energy that failed to be injected onto France’s electrical grid by renewable energy producers due to constraints.
2021 - 2022
Faculty Member at Université Paris Dauphine - PSL
Paris, France
Directed and led course on natural language processing for Master IASD. Supervised internships for master students.
2019 - 2020
AI R&D intern at NASA Ames Research Center
Mountain View, CA, USA
Joined NASA’s Planning & Scheduling group to conduct graph machine learning research and development on a hard scheduling problem known as Disjunctive Temporal Networks with Uncertainty (DTNUs) faced by NASA’s unmanned aerial vehicles. Advisors: Jeremy Frank, J. Benton.
• Created a time-based form of controllability for DTNUs and a tree search algorithm.
• Implemented a heuristic for tree search guidance based on graph neural networks.
• Carried out experiments in which the proposed approach significantly outperformed
state-of-the-art works, and remains the most efficient approach to date.
• Research contributed to solving scheduling problems faced by NASA’s UAVs.
2017 - 2021
Research Scientist at Safran
Paris, France
Research and development of learning-assisted planning systems for the company’s autonomous vehicles. Involves optimization of computations for NP-hard problems leveraging graph machine learning. Advisors: T. Cazenave, E. Jacopin, C. Guettier.
• Designed multiple approaches leveraging machine learning for more efficient path-planning under constraints, leading to substantial gains in planning performance.
• Implemented works in SAFRAN’s path-planning systems and supervised an intern.
• Published and presented articles at conferences describing works and filed a patent.
2016 - 2016
AI R&D intern at Osaka Prefecture University - RoboCup Laboratory
Osaka, Japan
Development of a (2D) RoboCup agent policy using machine learning and reinforcement learning. Worked with university’s RoboCup Team. Advisor: T. Nakashima.
• Trained a neural network to mimic agents’ policy. Applied reinforcement learning to the new agent. The created ML-based agent mimicked teacher agents fairly well.
Achievements:
-Extracted policy data from hand-coded RoboCup agents
-Trained a neural network in a supervised-learning fashion to mimic agents’ policy and implemented it inside a new agent
-Applied reinforcement learning to the new agent