Michael Figurnov
Also known as Михаил Фигурнов
Michael Figurnov is a senior staff research scientist at Google DeepMind, where he works on pretraining Gemini models. His doctoral work in computer science at the Higher School of Economics, supervised by Dmitry Vetrov, ran from 2013 to 2019. He designed the Evoformer architecture for AlphaFold 2, and he co-authored the Nature paper describing AlphaFold in 2021. His research papers include a CVPR 2017 paper on spatially adaptive computation, a 2018 arXiv paper on implicit reparameterization gradients, and a 2020 survey of Monte Carlo gradient estimation. He co-authored AlphaFold 3, which Nature published in May 2024.
Timeline
-
2013–2019
A PhD under Dmitry Vetrov
He studies for a PhD in computer science at the Higher School of Economics, supervised by Dmitry Vetrov.
-
2017
Spatially adaptive computation time
At CVPR 2017, he and six co-authors publish Spatially Adaptive Computation Time for Residual Networks.
-
May 2018
Implicit reparameterization gradients
In May 2018 he and two co-authors post Implicit Reparameterization Gradients on arXiv, a paper for NeurIPS 2018.
-
2020
A survey of Monte Carlo gradients
In 2020 he and three co-authors publish a broad survey of Monte Carlo gradient estimation in machine learning.
-
July 2021
Designing the Evoformer architecture
In July 2021 he is a co-author of the Nature paper describing AlphaFold, and he designs the Evoformer architecture for AlphaFold 2.
-
2022–2025
Protein modelling and LLMs for biology
From 2022 to 2025 his research covers protein modelling and LLMs for biology.
-
May 2024
Co-authoring AlphaFold 3
In May 2024 Nature publishes the AlphaFold 3 paper, and he is one of its co-authors.
-
Since January 2026
Pretraining Gemini models
In January 2026 he starts work on pretraining Gemini models at Google DeepMind, where he is a senior staff research scientist.
Sources and credits
Photo: We have no published photo of Michael Figurnov yet. Send us a free-licensed photo.