Projects
Academic and Professional Background
Julien Perez graduated from EPITA in 2005. He then earned a master’s degree in applied mathematics from Université Paris-Dauphine in 2006, followed by a Ph.D. in machine learning from Université Paris-Sud in 2010. His dissertation, supervised by Cécile Germain Renaud and Balázs Kégl within the TAO (Theme: Learning and Optimization) group of the Laboratoire de Recherche en Informatique (LRI), focused on non-Markovian decision processes and deep memory recurrent neural networks.
After completing his Ph.D., he undertook a postdoctoral fellowship at Télécom ParisTech, and later became a lecturer at Université Paris-Est Créteil. In 2013, he joined Naver Labs Europe (formerly Xerox Research Centre Europe), where he led research on deep learning, reinforcement learning, automatic reading, and dialogue systems.
Since March 2025, he is holding an Habilitation à Diriger des Recherches, "Contribution à la decision différentiable, raisonnement et sens commun". His current research involve generative model alignement and safety with application to pedagogy.
Activities at EPITA
Since returning to EPITA, Julien Perez has been actively involved in teaching artificial intelligence and machine learning. Notably, he introduced an innovative course on the fundamentals of deep learning, where students reimplement key concepts from scratch, including automatic differentiation, the design of dynamic computation graphs, and model optimization. He is also interested in agentic AI, an approach in which AI systems are capable of planning, reasoning, and acting autonomously.
Scientific Contributions
Julien Perez is the author of numerous scientific publications, particularly on memory-augmented neural networks, dialogue systems, and reinforcement learning.
Additional Resources
- Google Scholar profile: Julien Perez
- EPITA page: EPITA Research Laboratory