Welcome!
Pierre is a Phd researcher in artificial intelligence and neuroscience. His expertise spans from large model pretraining to invivo recording. Pierre defended his PhD in February 2025, in this website you will find a list of his published and current work along with recording of talks he presented. In January 2026, he started a series of weekly, short blog publications, aiming to share some insights he is gaining from his research.
Research interests
Why is it that every children go through similar stages when discovering language, math or music?
Theorizing the emergence of cognitive operations.
To theorize the emergence of cognitive algorithms like the detection of auditory structure, we replicate classical paradigms in the most recent artificial neural network architecture.
Modeling Development.
We model developmental patterns by probing the models throughout their pretraining. This experimental work has dedicated large compute ressources to measure the emergence of cognitive operations in neural network that just learn from looking at the data, like baby. From it, we observed a fundamental rule governing the learning dynamic: simple construct are learned before complex ones.
Predicting and recording brain activities.
Brain activity is predicted from model activity in response to the same stimulis.
Studying the representations of structures: from language to maths, in models and brains.
We propose novel approach to measure if linguistic structures are instantiated, in the brain or in artificial neural networks.