Scientific Profile

I did undergraduate studies in fundamental physics. Very soon, my interest became oriented toward the physics of the mesoscopic scale, where the collective dynamics of complex systems can lead to fascinating emergent phenomena. I therefore graduated in Complex Systems Physics to get a strong training in the techniques capturing such emergent properties: statistical physics, nonlinear dynamics, stochastic processes, etc…

Because the dynamics of neural networks seem to offer an ideal model system for such an approach, I did my Msc thesis in a computational neuroscience laboratory and I finalised my university training with a Msc in Cognitive Science in the Theoretical Neuroscience speciality.

I did the PhD in the group of Alain Destexhe. There, I benefited from a strong training in Computational Neuroscience. During the PhD, I developed new analytical tools to capture the dynamics of recurrent networks and to describe single neuron integration. Importantly I was also exposed to experimental neuroscience through both collaborations (team of F. Chavane in Univ. Aix-Marseille, M. V. Sanchez-Vives team in IDIBAPS Barcelona) and the interdisciplinary environment of the UNIC laboratory (CNRS, Gif/Yvette). This led me to perform the patch-clamp experiments testing the theoretical predictions of our modelling work (supervised by G. Ouanounou and T. Bal).

For my postdoctoral training, I wanted to strengthen my analytical skills as well as discover new aspects of neuroscientific research. I therefore joined the Neural Coding laboratory at the Italian Institute of Technology, an interdisciplinary initiative jointly led by Stefano Panzeri, a pioneer in the application of information theory in neural systems, and Tommaso Fellin, a specialist in optical and molecular engineering for interventional experiment in neuroscience.

I am currently finalising my postdoctoral training with a mixed experimental and theoretical project under the joint supervision of Alberto Bacci, a specialist of cortical microcircuits, and Nelson Rebola, a specialist of the cellular mechanisms of neuronal integration. Together, we study the contribution of inhibitory sub-populations of the cortex in visual processing.

In between my two postdocs (IIT and ICM), I helped the neuroinformatics group in setting up the data management pipeline of the EBrains platform for model sharing (the EU-funded platform for neuroscientific research, see https://ebrains.eu). This experience has made me familiar with the highest standards of data management and data sharing in modern research.

This interdisciplinary training combining theoretical and experimental expertise together with my diverse research experiences give me a unique scientific profile to tackle the challenging questions of modern neuroscience.

My research

As a general research line, my scientific work focuses on understanding how the dynamics of neural networks in the cerebral cortex can underlie the rich repertoire of computations performed by the brain of mammals.

More specifically, I currently investigate the functional role of the ongoing dynamics of network activity in shaping cortical computations. For example in the mouse primary visual cortex (the first cortical stage in the visual system), one observes a massive amount of visually-unrelated activity, why ? What is the functional role of this activity ? Surprinsingly, the variable levels of ongoing activity also have an impact on the neural representations of visual features. Neural activity patterns in the visual cortex following the same visual input can be variable depending on the level of ongoing activity.

This property is a challenge for our intuition about sensory systems. Cortical processing is not a fixed operation purely determined by the input, it rather seems to be a flexible process controlled by the ongoing dynamics of cortical networks.

The comparison with the current state-of-the-art of artificial intelligence nicely illustrate the non-trivial aspect of this property. In computer vision, the input-evoked activity represents 100% of the network activity in an artificial neural network trained to discriminate images. In contrast, the sensory-evoked activity is only a weak fraction of the network activity in the primary visual cortex of a mammal, and the evoked response shows a rather high variability over image presentations. The “state-dependent” nature of cortical computation is therefore a fascinating topic, whose function remains largely enigmatic.

Interestingly, previous research has shown that the behavioral context in which an animal stands is the main variable controling the dynamical regime of cortical network activity, thus potentially linking this functional flexibility to different behavioral needs.

This research line has two important applications. First a biomedical application. Indeed, characterizing neural computations in the healthy brain will allow to understand how complex psychiatric disorders, such as schizophrenia or depression, leads to cognitive dysfunction as a result of the (sometimes subtle) alterations in the network processing of information. Second, this drives advances in biologically-inspired artificial intelligence. In particular this research suggests specific algorithms to make artificial systems capable of the computational flexibility observed in the mammalian cortex.