Research

The machine learning group is a team of scientists working to build the next generation of machines that can safely and responsibly learn to act through interactions with humans, infrastructures, and other machines. We explore scientific breakthroughs in modern machine-learning technologies and real-world applications that can transform industries, improve people's lives, and benefit humanity. Our research focuses on:

• building trustworthy machine learning models with real-world data for several applications, such as medical AI and Robotics. Our approach involves Uncertainty Estimation, Computer Vision, and Explainable AI.

• pushing the boundaries of Deep Reinforcement Learning by developing models that can learn complex tasks and adapt and generalize across diverse environments. I envision a future where AI systems seamlessly transfer knowledge between tasks, enabling more robust, efficient, and scalable real-world applications.

• advancing Natural Language Processing to develop human-centered models, with a focus on personalized education, emotion recognition in conversational systems, and multilingual representation learning for low-resource languages. 

• building the theoretical foundations for collective and social intelligence involving multiple cooperative, competitive, or mixed-motive artificial agents, humans, and organizations. Our research has formalized centralized planning for the decentralized control paradigm, which provides an elegant and comprehensive framework for reasoning in ecologies of interacting artificial agents with imperfect information. This framework underpins almost all real-world interactions between humans and artificial agents.

This research is funded by government agencies nationally and internationally, see  ANR Projects: PLASMA (closed), DELICIO  (ongoing), MAESTRIOT  (ongoing), EpiRL (ongoing).