Alban Desmaison
Alban Desmaison is a research engineer at Meta who works on PyTorch, a deep-learning framework. He co-authored research on adaptive neural compilation in 2016 and the 2019 paper describing PyTorch's imperative, Pythonic design. After studying engineering and doing a PhD in machine learning and optimization, he maintained PyTorch's core features, joined its Core Maintainers in 2024, and served as a PyTorch Conference program chair in 2025. As of October 2026, he is PyTorch's Lead Core Maintainer.
Timeline
-
2016
Adaptive neural compilation
In 2016, he coauthors Adaptive Neural Compilation, which learns efficient programs by adapting them to a target distribution of inputs.
-
2019
The PyTorch research paper
In 2019, he coauthors the PyTorch paper, which describes an imperative, Pythonic machine-learning library that stays efficient while supporting GPUs.
-
2021
Facebook PyTorch engineering
In 2021, he is a research engineer at Facebook after studying engineering and doing a PhD in machine learning and optimization. He maintains PyTorch features such as autograd, optim, and nn.
-
2024
Joining PyTorch maintainers
In 2024, he joins PyTorch’s Core Maintainers, the group with final authority over technical decisions in the pytorch/pytorch project.
-
2025
A PyTorch conference chair
In 2025, he serves as a PyTorch Conference program chair and joins a Meta-led technical deep dive on the framework. His PyTorch work spans his University of Oxford PhD and his role at Meta, where he focuses on core components and feature design.
-
As of October 2026
Lead core maintainer of PyTorch
As of October 2026, he is a Research Engineer at Meta and the Lead Core Maintainer of PyTorch. In May 2026, he joins a live Q&A on the PyTorch 2.12 release, and the August 2026 conference announcement lists him for a PyTorch Updates keynote.
Sources and credits
Photo: We have no published photo of Alban Desmaison yet. Send us a free-licensed photo.
- Adaptive Neural Compilation
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Accelerating PyTorch with CUDA Graphs
- Alban Desmaison and Piotr Bialecki are now PyTorch Core Maintainers
- PyTorch Conference Keynote Speakers
- Accelerating the Future of Open Source AI: PyTorch Conference 2025 Recap
- PyTorch 2.12 Release Live Q&A
- PyTorch Conference North America Announces 2026 Keynotes