PeopleofSuperintelligence

Researcher

Dominique Beaini

Dominique Beaini is an AI researcher whose work connects computer vision and robotics with graph learning for molecular science. He completed a robotics and computer vision PhD at Polytechnique Montréal, then coauthored Directional Graph Networks and later contributed to graph and molecular learning tools such as Graphium, MolGPS and SeedER. In 2025, he coauthored a Valence Labs perspective on virtual cells that predict cellular responses to perturbations and explain molecular effects. He leads a research unit at Valence Discovery and is an adjunct professor at Université de Montréal.

Now at
Valence Discovery and Université de Montréal
Badges

Timeline

  1. 2018–2019

    Electromagnetic kernels and a PhD

    In 2018, he coauthors a paper on convolution kernels based on electromagnetic potentials and fields for computer vision and shape analysis. In 2019, his doctoral thesis at Polytechnique Montréal examines Green Function and Electromagnetic Potential methods for computer vision and convolutional neural network applications. His doctoral research covers robotics and computer vision.

  2. 2020

    Directional Graph Networks

    In 2020, he coauthors Directional Graph Networks, which introduces globally consistent anisotropic kernels for graph convolutions. Its benchmarks show relative error reductions of 8% on CIFAR10 graph data and 11% to 32% on molecular ZINC.

  3. June 2021

    Graph research at Valence

    In June 2021, he gives a Cambridge AI Research Group talk on deep learning for graphs as Head of Graph Research at Valence Discovery. The company focuses on drug discovery through property prediction, screening and generation.

  4. 2023

    Graphium for molecular learning

    In 2023, he writes Graphcore's introduction to Graphium, an open-source library for graph representation learning on real-world chemistry tasks. Its API supports training large-scale molecular GNNs.

  5. 2024

    MolGPS for molecular graphs

    In 2024, he coauthors a study on scaling graph neural networks for molecular graphs that introduces MolGPS, a graph foundation model. MolGPS outperforms previous state-of-the-art models on 26 of 38 downstream tasks.

  6. 2025

    A virtual-cell framework

    In 2025, he coauthors Valence Labs' perspective on virtual cells, which should predict cellular responses to perturbations and explain them through changes to biomolecular interactions. It proposes a lab-in-the-loop approach and biologically grounded benchmarks to guide virtual-cell development.

  7. May 2026

    SeedER for knowledge graph retrieval

    In May 2026, he coauthors SeedER, a first-stage knowledge-graph retrieval method that uses Graph Transformers and reinforcement learning and can run on a CPU. The paper reports performance competitive with approaches fine-tuned from large language models.

  8. As of October 2026

    Research leadership at Valence

    As of October 2026, he leads a research unit at Valence Discovery and is an adjunct professor in Université de Montréal's Department of Computer Science and Operations Research. His research interests include graph neural networks, molecular modeling and drug discovery.

Career

  1. Since 2021Valence DiscoveryHead of Graph Research
  2. NowUniversité de MontréalAdjunct Professor

Education

  1. 2019Polytechnique MontréalPhD

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Université de Montréal. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Université de Montréal. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Université de Montréal. 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. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Université de Montréal. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.

Sources and credits

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

  1. Novel Convolution Kernels for Computer Vision and Shape Analysis based on Electromagnetism arxiv.org. Checked 22 October 2026.
  2. Documents dont l'auteur est "Beaini, Dominique" - PolyPublie publications.polymtl.ca. Checked 22 October 2026.
  3. Dominique Beaini | Mila mila.quebec. Checked 22 October 2026.
  4. Directional Graph Networks arxiv.org. Checked 22 October 2026.
  5. Unlocking Deep Learning for Graphs talks.cam.ac.uk. Checked 22 October 2026.
  6. Graphium: an IPU-ready Python Library for training molecular GNNs at scale graphcore.ai. Checked 22 October 2026.
  7. On the Scalability of GNNs for Molecular Graphs arxiv.org. Checked 22 October 2026.
  8. Virtual Cells: Predict, Explain, Discover arxiv.org. Checked 22 October 2026.
  9. Virtual Cells: Predict, Explain, Discover - Valence Labs valencelabs.com. Checked 22 October 2026.
  10. SeedER: Seed-Expand-Retrieve for Efficient Knowledge Graph Retrieval arxiv.org. Checked 22 October 2026.