PeopleofSuperintelligence

Researcher

Eugene Belilovsky

Eugene Belilovsky is a machine-learning researcher whose work develops efficient methods for deep learning and computer vision. He completed a joint PhD at CentraleSupélec, Paris-Saclay and KU Leuven in 2018, then coauthored layerwise CNN training methods that scaled to ImageNet. His work also spans embodied question answering, text-to-3D generation and federated learning. Since 2021, he has been a professor at Concordia and an academic member of Mila; he received tenure in 2025, and his group had four papers accepted at ICML in 2026.

Born
1987

Years link to their sources.

Timeline

  1. 2018

    A joint doctoral degree

    In 2018, he completes a joint PhD at CentraleSupélec, University of Paris-Saclay and KU Leuven (VISICS), supervised by Matthew Blaschko.

  2. 2018

    Embodied question answering

    In 2018, he coauthors Blindfold Baselines for Embodied QA. The paper reports that a question-only baseline achieves state-of-the-art results on the EmbodiedQA task except when the agent is spawned extremely close to the object.

  3. 2019

    Layerwise CNN training

    In 2019, he coauthors Greedy Layerwise Learning Can Scale To ImageNet, which builds deep networks layer by layer and trains a VGG-11 model to match end-to-end learning accuracy.

  4. 2020

    Decoupled CNN training

    In 2020, he coauthors Decoupled Greedy Learning of CNNs with Michael Eickenberg and Edouard Oyallon. The method separates training across layers and supports asynchronous modules through replay buffers, with experiments on CIFAR-10 and ImageNet.

  5. Since 2021

    Concordia professor and Mila member

    Since 2021, he is a professor at Concordia University and an academic member of Mila. Before this appointment, he works as a postdoctoral researcher at Mila and the University of Montreal with Aaron Courville.

  6. 2022–2023

    Text-to-3D and federated learning

    In 2022, he coauthors CLIP-Mesh, which generates a 3D model from a text prompt without 3D supervision. In 2023, he coauthors a federated-learning study that uses pretrained models and a Nearest Class Means classifier for computer-vision tasks.

  7. 2025

    Tenure in 2025

    In 2025, he receives tenure.

  8. As of October 2026

    A professor and four ICML papers

    As of October 2026, he is a professor at Concordia University and an associate academic member of Mila. In July 2026, his group has four papers accepted at ICML 2026 on local optimization, efficient architectures, model merging and optimization.

Career

  1. Since 2021MilaPostdoctoral Researcher
  2. Since 2021Concordia UniversityProfessor
  3. Université de MontréalPostdoctoral Researcher

Education

  1. 2018Université Paris-SaclayPhD
  2. 2018Université Paris-SaclayPhD
  3. 2018KU LeuvenPhD

From the biography.

  1. Universities (31%) and startups and other companies (25%) employ the largest shares of 2020–2026 entrantsIncludes Mila. Among 819 people whose first recorded role started in 2020–2026 and who have a known current organization, 60% work in commercial sectors (25% at startups and other companies, 23% at frontier labs and 13% at big tech, with 9 in two), and 31% work at universities, compared with 66% for earlier entrants.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Mila. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Mila. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Mila. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Eugene Belilovsky yet. Send us a free-licensed photo.

  1. About | Eugene Belilovsky eugenium.github.io. Checked 21 October 2026.
  2. Blindfold Baselines for Embodied QA arxiv.org. Checked 21 October 2026.
  3. Greedy Layerwise Learning Can Scale To ImageNet proceedings.mlr.press. Checked 21 October 2026.
  4. Decoupled Greedy Learning of CNNs proceedings.mlr.press. Checked 21 October 2026.
  5. CLIP-Mesh: Generating textured meshes from text using pretrained image-text models arxiv.org. Checked 21 October 2026.
  6. Guiding The Last Layer in Federated Learning with Pre-Trained Models arxiv.org. Checked 21 October 2026.
  7. Eugene Belilovsky | Mila mila.quebec. Checked 21 October 2026.