Grégoire Mialon

Postdoctoral Researcher

Meta AI

About me

I am a Postdoctoral Researcher at Meta AI in Paris, where I am advised by Yann LeCun. My research currently focuses on self-supervised learning. I am also curious about technology and its implications in politics and entrepreneurship. I sometimes write on these subjects for different organizations such as the Group for Geopolitical Studies, a european think-tank hosted at École normale supérieure.

Prior to that, I was a PhD Student in Machine Learning in the Sierra research team at Inria, where I was co-advised by Julien Mairal and Alexandre d'Aspremont. During my PhD, I proposed architectural improvements for learning from small or medium sized datasets such as scientific data.


  • Self-supervised learning
  • Kernel methods
  • Machine learning for Fundamental Science
  • Technology policies and Entrepreneurship


  • PhD in Machine Learning, 2022


  • MSc in Machine Learning (MVA), 2018

    ENS Paris-Saclay

  • Diplôme d'Ingénieur (Theoretical Physics, Business Studies), 2017

    École polytechnique

  • Research Intern, 2016

    MIT, Department of Nuclear Science and Engineering


All news»

[07/06/22] I talked about Graph Transformers at X-IA#10: State of the Graphs in Paris and at SGDMAL in Rennes.

[21/02/22] I started to work at Meta AI as a Postdoctoral Researcher.

[19/01/22] I successfully defended my PhD, whose title is “On Inductive Biases for Machine Learning in Data Constrained Settings”.

[04/02/21] I talked about “How Natural Language Processing is reshaping Machine Learning” at the Paris NLP Meetup.

[12/01/21] Our paper A Trainable Optimal Transport Embedding for Feature Aggregation and its Relationship to Attention was accepted to ICLR 2021.

Recent & Upcoming Talks

Recent Posts

Notes I wrote (in French) for various organizations.

L'oeil d'Apple dans nos iPhones ?

Les outils développés par Apple pour traquer et signaler automatiquement les images pédopornographiques stockées sur les smartphones de ses utilisateurs soulèvent de nombreuses inquiétudes relatives à la protection des données. Ils sont surtout le symptôme d'une évolution dans les standards de traitement des données par les géants du numérique. Une analyse à retrouver sur le site de la revue Esprit.

Le logiciel libre, outil de soft power et de lutte contre le Covid-19

Gustave Ronteix et moi-même avons interrogé Alexandre Gramfort, chercheur en machine learning à Inria Saclay, pour le Grand Continent. Open source comme outil stratégique, naissance de scikit-learn, IA en santé, GAFAs et monde académique : le résultat est passionnant et disponible ici.