After bachelors in biology and philosophy, my journey in neuroscience started at École Normale Supérieure (Paris), where I first studied the intracellular pathways activated by atypical antipsychotics in mice. After that, I worked for seven years with humans on various topics spanning social, cognitive and computational neurosciences, until it became obvious that invasive system neuroscience techniques were also necessary to answer my questions and fulfill my curiosity regarding the information broadcasted by neuromodulators and the pathways mediating their multifaceted influence on behavior.

​

Obviously, integrating human and rodent data within a cohesive framework involves a number of methodological and conceptual challenges. But I believe that this effort is necessary to meet the expectations of clinicians and societies with respect to neuroscience as a whole. Because they have been shaped by the same evolutionary mechanisms to fulfill the same function, many fundamental principles ought to be conserved across species, especially when more than 85% of their coding DNA can be aligned. Beyond the specific hypotheses that I plan to tests in the next few years, my goal is to push this cross-species integration to the limit with the help of computational models (i) to design experiments probing species-independent mechanisms and (ii) to account for species-dependent processes when analyzing the data.

​

In sum, I try to wear cognitive, system and computational neuroscience hats altogether —which may look funny at times— to animate and fuel a multidisciplinary group that will hopefully constitute a fertile environment and a stimulating playground for my coworkers, at the frontier of the unknown and beyond.

​

Nowadays, the frontier that fascinates me the most has to do with the paradoxical interplay of causality and purpose in modern neuroscientific theories. Indeed, science not longer accepts the old Aristotelian idea that purposes can be treated as causes. And yet, prominent frameworks such as reinforcement-learning and predictive coding —that treat the brain as a controller— derive brain activities and behavior as whole from the definition of "fundamental" unifying purposes, be it utility-maximization or uncertainty-minimization. This consideration explains why I never felt compelled to choose a side between these two frameworks that are equally inspiring and incredibly complementary when it comes to modeling brain activity and behaviors. And it also explains why I envision seriously the possibility of other "fundamental" purposes, such as controllability maximization or empowerment, alongside (not to say above) these two! Finally, my interest in neuromodulatory systems stems directly from this hypothetical (and paradoxical) plurality of purposes, as these highly conserved systems are very well suited to signal a variety of control errors.