AI People Wiki

Builders & open source

Samuli Laine

Member of NVIDIA Research since 2007, NVIDIA

Samuli Laine joined NVIDIA Research in 2007, and NVIDIA's page says Laine's research has ranged from photorealistic and real-time rendering to voxel-based graphics, GPU ray tracing, and 3D content creation, with a current focus on deep learning and its applications in computer graphics. Laine is first author of "Modular Primitives for High-Performance Differentiable Rendering", the paper behind NVIDIA's nvdiffrast PyTorch library. Laine is also a co-author of the StyleGAN, StyleGAN3, EDM, and EDM2 generative model papers, whose official implementations NVIDIA publishes in its NVlabs GitHub organization.12345678

At a glance

Current role
Member of NVIDIA Research since 2007
Company
NVIDIA
Sector
Company
Work listed
2
Sources cited
12
Profile reviewed

Career

Samuli Laine joined NVIDIA Research in 2007 after completing a PhD at Helsinki University of Technology. NVIDIA's profile says Laine's research has ranged from photorealistic and real-time rendering to voxel-based graphics, GPU ray tracing and 3D content creation, and that Laine now focuses on deep learning and its applications in computer graphics.1

Research

With Timo Aila, Laine wrote "Temporal Ensembling for Semi-Supervised Learning" (2016), and Laine was a co-author of Noise2Noise (2018), which trains image restoration networks without clean target images. Laine is also first author of "Modular Primitives for High-Performance Differentiable Rendering" (2020), the paper behind NVIDIA's nvdiffrast, a PyTorch library of primitive operations for rasterization-based differentiable rendering.91023

Laine has co-authored a line of image generation papers with Tero Karras and colleagues: progressive growing of GANs (2017), the style-based generator StyleGAN (2018), StyleGAN2 (2019) and the alias-free StyleGAN3 (2021). In diffusion models, Laine co-authored EDM (2022), which maps out the design space of diffusion-based generative models, and EDM2 (2023), which analyzes and improves their training dynamics; the official EDM2 implementation is published in NVIDIA's NVlabs GitHub organization.114125678

Work and contributions

  • nvdiffrast Published research

    First author of "Modular Primitives for High-Performance Differentiable Rendering" (2020). The nvdiffrast repository, a PyTorch library of primitive operations for rasterization-based differentiable rendering, lists Laine first and is made available under the NVIDIA Source Code License.23

  • StyleGAN and EDM generative models Published research

    Second author of the 2018 StyleGAN paper and co-author of StyleGAN3 (2021), EDM (2022), and EDM2 (2023) with Tero Karras, Miika Aittala, Timo Aila, and others. The EDM2 repository releases code and pretrained models under CC BY-NC-SA 4.0.45678

Timeline

  1. Joined NVIDIA Research after a PhD at Helsinki University of Technology.1
  2. Co-authored "Temporal Ensembling for Semi-Supervised Learning" with Timo Aila.9
  3. Co-authored "Progressive Growing of GANs for Improved Quality, Stability, and Variation".11
  4. Co-authored "Noise2Noise: Learning Image Restoration without Clean Data".10
  5. Co-authored StyleGAN, a style-based generator architecture for GANs.4
  6. Co-authored StyleGAN2, "Analyzing and Improving the Image Quality of StyleGAN".12
  7. First author of the nvdiffrast paper on modular primitives for differentiable rendering.2
  8. Co-authored "Alias-Free Generative Adversarial Networks" (StyleGAN3).5
  9. Co-authored EDM, "Elucidating the Design Space of Diffusion-Based Generative Models".6
  10. Co-authored EDM2, on the training dynamics of diffusion models.7

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

12 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Samuli Laine — NVIDIA ResearchNVIDIA Research · Official page · accessed Oct 21, 2026
  2. Modular Primitives for High-Performance Differentiable Rendering (arXiv 2011.03277)arXiv · Paper · published Nov 6, 2020 · accessed Oct 21, 2026
  3. NVlabs/nvdiffrast README (NVIDIA Source Code License)NVlabs (GitHub repository) · Repository · accessed Oct 21, 2026
  4. A Style-Based Generator Architecture for Generative Adversarial Networks (arXiv 1812.04948)arXiv · Paper · published Dec 12, 2018 · accessed Oct 21, 2026
  5. Alias-Free Generative Adversarial Networks (arXiv 2106.12423)arXiv · Paper · published Jun 23, 2021 · accessed Oct 21, 2026
  6. Elucidating the Design Space of Diffusion-Based Generative Models (arXiv 2206.00364)arXiv · Paper · published Jun 1, 2022 · accessed Oct 21, 2026
  7. Analyzing and Improving the Training Dynamics of Diffusion Models (arXiv 2312.02696)arXiv · Paper · published Dec 5, 2023 · accessed Oct 21, 2026
  8. NVlabs/edm2 README (EDM2 and Autoguidance; CC BY-NC-SA 4.0)NVlabs (GitHub repository) · Repository · accessed Oct 21, 2026
  9. Temporal Ensembling for Semi-Supervised Learning (arXiv 1610.02242)arXiv · Paper · published Oct 7, 2016 · accessed Oct 21, 2026
  10. Noise2Noise: Learning Image Restoration without Clean Data (arXiv 1803.04189)arXiv · Paper · published Mar 12, 2018 · accessed Oct 21, 2026
  11. Progressive Growing of GANs for Improved Quality, Stability, and Variation (arXiv 1710.10196)arXiv · Paper · published Oct 27, 2017 · accessed Oct 21, 2026
  12. Analyzing and Improving the Image Quality of StyleGAN (arXiv 1912.04958)arXiv · Paper · published Dec 3, 2019 · accessed Oct 21, 2026