Ekin Dogus Cubuk
Also known as Dogus Cubuk
Ekin Dogus Cubuk is a machine-learning researcher with a physics background who co-created AutoAugment, a procedure that automatically searches for improved data augmentation policies. He studied physics and engineering at Swarthmore College and then earned a PhD in applied physics at Harvard. Google Brain hired him in 2018, and his GNoME work with Amil Merchant reported 2.2 million new crystals in a 2023 Nature study. In 2025 he left Google to cofound Periodic Labs, which launched with a claimed $300 million seed round. He is co-CEO of the company, which builds LLM agents for physical sciences.
- Now at
- Periodic Labs
- Works
- 132
- Citations
- 24,242
- h-index
- 49
Publication numbers from OpenAlex, as of 15 October 2026.
Timeline
-
2006–2010
Physics and engineering at Swarthmore
He studies physics and engineering at Swarthmore College from 2006 through 2010. At Harvard, he completes a PhD in applied physics.
-
2018
Joining Google Brain
Google Brain hires him in 2018 after a one-year residency. He later leads the chemistry and materials science team at DeepMind.
-
2018–2019
The AutoAugment method
He and Barret Zoph post AutoAugment on the Google AI Blog in June 2018. The CVPR 2019 paper describes a simple procedure to automatically search for improved data augmentation policies.
-
2023
GNoME and 2.2 million crystals
He and Amil Merchant report GNoME in Nature in 2023. The system finds 2.2 million new crystals, including 380,000 stable materials.
-
2025
Cofounding Periodic Labs
He leaves Google in 2025 to cofound Periodic Labs. The company launches in September 2025 with a claimed $300 million seed round.
-
As of October 2026
Co-CEO of Periodic Labs
As of October 2026, he is co-CEO of Periodic Labs, which builds LLM agents for physical sciences with physical experiments in the loop.
Sources and credits
Photo: We have no published photo of Ekin Dogus Cubuk yet. Send us a free-licensed photo.
- Ekin Dogus Cubuk | nextomoro
- Ekin Dogus Cubuk runs a startup to accelerate physics R&D using AI
- Speaker Details: RAISE Summit 2026
- Improving Deep Learning Performance with AutoAugment
- AutoAugment: Learning Augmentation Strategies From Data
- Scaling deep learning for materials discovery
- Millions of new materials discovered with deep learning
- Periodic Labs launches with $300M to build an “AI scientist”