PhD student in computer science at Inria Rennes, working on the detection of AI-manipulated content.
In 2025 I obtained my master's degree in computer science as well as the agrégation (French national competitive examination), and I have been a PhD student at Inria Rennes since September 2025.
My research focuses on the security of artificial intelligence — more specifically, on the detection of deepfakes and content manipulated by generative models.
Artificial intelligence models play an ever-growing role in everyday life, often without us noticing. This is particularly the case for AI-generated or AI-modified photo and video content, whose spread raises serious issues — especially on social media, in the context of disinformation.
New generative models appear at a steady pace : it is essential to design detection methods that do not depend on the models seen during training, and that generalise to the latest architectures. One promising approach is to train a detector with no manipulated examples (zero-shot), or with very few of them (few-shot) — this is the direction my thesis aims to explore.
IRISA — University of Rennes
French national competitive examination for secondary-education teachers — rank 15/22
École Normale Supérieure de Rennes / University of Rennes
École Normale Supérieure de Rennes — Bachelor's (L3, mathematics minor) then Master's (SIF)
Lycée Albert Schweitzer — Scientific track, MPSI / MP*
Lycée André Boulloche — Highest Honours
As part of my PhD, I teach computer science to undergraduate students.
Tutorial BSc, Year 3 · ISTIC
Tutorial BSc, Year 3 · ISTIC
End-of-studies internship in the Artishau team at Inria Rennes, supervised by Ewa Kijak. The work focused on detecting photo and video content manipulated by AI, with the goal of generalising to generative models unseen during training (zero-shot settings).
Internship at Euronext in Porto, focused on the application of European cybersecurity regulation within a market infrastructure.
Under the supervision of Teddy Furron at the IRISA laboratory in Rennes, I worked on the creation of adversarial images for AI classification models, in order to study their robustness against imperceptible perturbations.