Timo Aila

Timeline

1. 2005–2007

   From Hybrid Graphics to NVIDIA

   He completes an article-based doctoral thesis, Efficient Algorithms for Occlusion Culling and Shadows, at Helsinki University of Technology in 2005. His expertise spans real-time game rendering, from Max Payne to dPVS, the first commercial occlusion culling library, and he is chief scientist of Hybrid Graphics when NVIDIA acquires it in 2006. In 2007, he joins NVIDIA Research after leading the computer graphics research group at Helsinki University of Technology.

2. 2009

   Ray traversal on GPUs

   In 2009, he and Samuli Laine publish Understanding the Efficiency of Ray Traversal on GPUs at High-Performance Graphics, which narrows the gap between simulated and measured speed and yields the fastest GPU ray tracer to date. He also has a central role in NVIDIA's research on ray tracing, including the design of the RTX hardware units.

3. 2017–2018

   Noise2Noise image restoration

   He contributes to a SIGGRAPH 2017 paper on interactive reconstruction of Monte Carlo image sequences with a recurrent denoising autoencoder. In 2018, he coauthors Noise2Noise with Jaakko Lehtinen, Samuli Laine, Tero Karras and others at ICML, showing that a model can learn to restore images from corrupted examples alone, including photographic noise, Monte Carlo renders and undersampled MRI scans.

4. 2018–2019

   Progressive GANs and StyleGAN

   In 2018, he coauthors Progressive Growing of GANs with Tero Karras, Samuli Laine and Jaakko Lehtinen, an ICLR oral presentation that grows the generator and discriminator layer by layer from a low resolution. The 2019 CVPR paper A Style-Based Generator Architecture for Generative Adversarial Networks, written with Karras and Laine, introduces the generator known as StyleGAN and receives a best paper honorary mention as an oral presentation.

5. 2022

   A NeurIPS Outstanding Paper

   In 2022, he coauthors Elucidating the Design Space of Diffusion-Based Generative Models with Tero Karras, Miika Aittala and Samuli Laine, and the paper receives an Outstanding Paper award at NeurIPS. It separates the concrete design choices of diffusion models and reaches a state-of-the-art FID of 1.79 on class-conditional CIFAR-10 with 35 network evaluations per image.

6. 2024

   NeurIPS best paper runner-up

   In 2024, he coauthors Guiding a Diffusion Model with a Bad Version of Itself with Tero Karras, Miika Aittala, Tuomas Kynkäänniemi, Jaakko Lehtinen and Samuli Laine, an oral presentation named a Best Paper Runner-up at NeurIPS 2024. The method guides generation with a smaller, less-trained version of the model itself and sets record FIDs of 1.01 for ImageNet at 64x64 and 1.25 at 512x512.

7. 2025

   A Technical Achievement Award

   He wins a Technical Achievement Award from the Academy of Motion Picture Arts and Sciences in 2025, shared with Attila T. Áfra, for contributions to AI image denoising. He led early work applying the U-Net architecture to denoising, a technique that became vital across the film industry.

8. As of October 2026

   Distinguished scientist at NVIDIA

   As of October 2026, he is a Senior Distinguished Research Scientist at NVIDIA in Helsinki and a docent at Aalto University. His research sits at the intersection of neural network architectures, machine learning and computer graphics, with a particular focus on generative modeling.