PeopleofSuperintelligence

Researcher

Mathis Petrovich

Mathis Petrovich is a research scientist at NVIDIA who builds generative models that turn natural language into 3D human motion. After studying computer science at École normale supérieure Paris-Saclay, he completed an ELLIS PhD in 2024 in the IMAGINE team at École des Ponts ParisTech, working with the Max Planck Institute for Intelligent Systems under Gül Varol and Michael J. Black. During the doctorate he developed ACTOR for action-conditioned motion, TEMOS for diverse text-to-motion synthesis, TMR for text-to-motion retrieval and STMC for multi-track timeline control, and his thesis was honoured at the Paris-Est Sup 2024 Thesis Awards in December 2025. In NVIDIA's Data-Driven AI for Robotics group, he co-authored Kimodo, a motion diffusion model for 3D human and robot motion, in March 2026, and he works on text-to-3D human motion synthesis.

Now at
NVIDIA
Badges

Timeline

  1. By 2021

    An ELLIS PhD in two labs

    After a BS in computer science and the MVA master's degree at École normale supérieure Paris-Saclay, he becomes an ELLIS PhD student in the IMAGINE computer vision team at École des Ponts ParisTech and the Perceiving Systems Department of the Max Planck Institute for Intelligent Systems. Gül Varol and Michael J. Black co-advise him, and he presents his first research as a poster at the ELLIS Doctoral Symposium in 2021.

  2. 2021

    ACTOR for action-conditioned motion

    With Michael J. Black and Gül Varol, he introduces ACTOR at ICCV 2021, a Transformer-based variational autoencoder for human motion. The model synthesizes variable-length sequences of SMPL body models conditioned on a categorical action, and its synthetic data helps train action recognition.

  3. 2022–2023

    TEMOS and TMR

    TEMOS, his ECCV 2022 model, pairs a motion variational autoencoder with a text encoder to generate multiple diverse 3D human motions from one textual description. At ICCV 2023, he presents TMR, which extends TEMOS with a contrastive loss for text-to-motion retrieval and, in one comparison, lowers the median rank from 54 to 19.

  4. 2024

    Timeline control and a doctorate

    At the CVPR 2024 HuMoGen workshop, he presents STMC, a method for multi-track timeline control of text-driven 3D human motion, with coauthors including Umar Iqbal, Xue Bin Peng and Davis Rempe. He completes his PhD in the IMAGINE team in 2024, with a thesis on natural language control for 3D human motion synthesis.

  5. By 2025

    Research scientist at NVIDIA

    By June 2025, he returns to NVIDIA, where he had interned in Sanja Fidler's team, as a research scientist working on 3D human motion synthesis. He says he chose industry for the chance to build a project from top to bottom and see it become a usable product.

  6. December 2025

    A Paris-Est Sup thesis award

    In December 2025, his thesis "Natural Language Control for 3D Human Motion Synthesis", completed at LIGM-Imagine under Gül Varol and Michael J. Black, is honoured at the Paris-Est Sup 2024 Thesis Awards ceremony.

  7. March 2026

    The Kimodo motion model

    In March 2026, he is the second author of Kimodo, a 24-author paper on a kinematic motion diffusion model trained on 700 hours of optical motion capture data. The model generates 3D human and robot motions from text prompts and kinematic constraints. He works on its model team under project lead Davis Rempe.

  8. As of October 2026

    Research scientist in NVIDIA DAIR

    He is a research scientist in NVIDIA Research's Data-Driven AI for Robotics (DAIR) group, led by Umar Iqbal, which studies how robots can learn from human videos, motion capture and demonstrations. In July 2026, he co-authors ARDY and MotionBricks, two motion generation papers in ACM Transactions on Graphics (SIGGRAPH).

Career

  1. NowNVIDIAIntern

Education

  1. From 2021École des Ponts ParisTechPhD student
  2. From 2021Max Planck Institute for Intelligent SystemsPhD student
  3. École Normale Supérieure Paris-SaclayBS

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Mathis Petrovich. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. Google has the highest combined h-index among AI organizationsIncludes NVIDIA. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes NVIDIA. 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.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes NVIDIA. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Mathis Petrovich | ELLIS Institute Tübingen tue.ellis.eu. Checked 20 October 2026.
  2. Beyond the Thesis with Mathis Petrovich | European Laboratory for Learning and Intelligent Systems ellis.eu. Checked 20 October 2026.
  3. [2104.05670] Action-Conditioned 3D Human Motion Synthesis with Transformer VAE arxiv.org. Checked 20 October 2026.
  4. [2204.14109] TEMOS: Generating diverse human motions from textual descriptions arxiv.org. Checked 20 October 2026.
  5. [2305.00976] TMR: Text-to-Motion Retrieval Using Contrastive 3D Human Motion Synthesis arxiv.org. Checked 20 October 2026.
  6. [2401.08559] Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation arxiv.org. Checked 20 October 2026.
  7. Computer Vision News - June 2024 – Page 34 rsipvision.com. Checked 20 October 2026.
  8. Three doctors from the School awarded Thesis Prizes | École nationale des ponts et chaussées ecoledesponts.fr. Checked 20 October 2026.
  9. [2603.15546] Kimodo: Scaling Controllable Human Motion Generation arxiv.org. Checked 20 October 2026.
  10. Kimodo: Scaling Controllable Human Motion Generation research.nvidia.com. Checked 20 October 2026.
  11. DAIR @ NVIDIA Research research.nvidia.com. Checked 20 October 2026.
  12. Mathis Petrovich | DAIR @ NVIDIA Research research.nvidia.com. Checked 20 October 2026.