Ekin Dogus Cubuk

Ekin Dogus Cubuk is a Turkish-American physicist and machine-learning researcher. He is the co-founder of Periodic Labs, the 2025 San Francisco startup that pairs AI scientists with autonomous robotic laboratories to accelerate materials discovery, alongside Liam Fedus as fellow co-founder. He was previously a senior research scientist at Google DeepMind, where he led the chemistry, physics, and materials-science research group and was a senior author on the GNoME materials-discovery line, after joining Google Brain in 2017 from a postdoc at Stanford and a doctorate in applied physics at Harvard.

At a glance

• Education: BA in physics and BS in engineering, Swarthmore College (2010); PhD in applied physics, Harvard University (2016), under Efthimios Kaxiras; postdoctoral scholar at Stanford University Materials Science and Engineering (2016 to 2017).
• Current role: Co-founder of Periodic Labs since March 2025.
• Key contributions: Co-author of AutoAugment and RandAugment, the data-augmentation methods that became standard practice in deep-learning training pipelines; senior author of the GNoME materials-discovery paper in Nature (Merchant et al., 2023) reporting 2.2 million predicted stable crystal structures; chemistry, physics, and materials-science research lead at Google DeepMind through 2025; co-founder of Periodic Labs with Liam Fedus in March 2025.
• X / Twitter: @ekindogus
• LinkedIn: Ekin Dogus Cubuk
• Google Scholar: Ekin Dogus Cubuk
• arXiv: Cubuk, Ekin Dogus

Origins

Cubuk is Turkish-American. He completed undergraduate study at Swarthmore College in Pennsylvania from 2006 through 2010, earning a BA in physics with high honors and a BS in engineering. He completed his PhD in applied physics at Harvard University in 2016, under Efthimios Kaxiras, Professor of Physics and Applied Physics. The doctoral research addressed structural complexities in atomistic systems, ranging from disordered solids and supercooled liquids to catalytic surfaces and lithium-ion battery materials, combining first-principles density-functional-theory calculations with machine learning and stochastic simulation. The dissertation was an early instance of the ML-plus-physics methodology that anchored his later career.

Career

After Harvard, Cubuk completed a postdoctoral year at Stanford University in the Materials Computation and Theory Group led by Evan Reed, working on the screening of solid lithium-ion conductor materials and on transfer-learning approaches to small materials-science datasets.

In 2017 he joined Google Brain through the Google AI Residency program, transitioning to a full research scientist position in 2018. The early Google Brain period produced the data-augmentation papers that became foundational training-pipeline contributions. AutoAugment (Cubuk, Zoph, Mane, Vasudevan, Le, 2018) introduced reinforcement-learning-based search over augmentation policies and achieved state-of-the-art image-classification accuracy. RandAugment (Cubuk, Zoph, Shlens, Le, 2019) simplified the same idea into a hyperparameter-light alternative that became standard across vision-model training. SpecAugment (Park, Chan, Zhang, Chiu, Zoph, Cubuk, Le, 2019) extended the methodology to automatic speech recognition. The mid-Google period extended his research into semi-supervised learning and model robustness, producing widely-cited collaborations including FixMatch (Sohn et al., 2020) and AugMix (Hendrycks et al., 2019).

Following the April 2023 merger of Google Brain and DeepMind into Google DeepMind, Cubuk's group transitioned into the consolidated organization. He was promoted to senior research scientist and built out the materials-science research line that culminated in the GNoME project. The November 2023 Nature paper (Merchant, Batzner, Schoenholz, Aykol, Cheon, Cubuk, 2023), with Cubuk as senior author, introduced Graph Networks for Materials Exploration and reported 2.2 million predicted stable inorganic crystal structures, approximately 380,000 of them estimated stable at zero kelvin and therefore plausible candidates for experimental synthesis. The paper appeared alongside the autonomous-laboratory paper by Szymanski et al. (2023), on which Cubuk was also listed; the two papers framed the closed-loop AI-prediction-and-robotic-synthesis methodology that became the Periodic Labs founding thesis. By 2025 he led approximately 22 researchers at Google DeepMind covering chemistry, physics, materials science, and the underlying machine-learning methodology for scientific discovery, having spent eight years at Google in total.

In March 2025 Cubuk co-founded Periodic Labs with Liam Fedus, who had departed OpenAI the same month. The founding thesis combined Fedus's frontier post-training methodology with Cubuk's materials-discovery and autonomous-laboratory background, organized around a closed-loop architecture pairing AI hypothesis generation with robotic experiment execution. The pair had met at Google Brain approximately seven years before the founding. In September 2025 Periodic Labs emerged from stealth with a $300 million seed round at a $1.3 billion post-money valuation led by Andreessen Horowitz and Felicis, plus DST Global, NVIDIA's NVentures arm, Accel, Jeff Bezos, Eric Schmidt, Elad Gil, and Jeff Dean. As of March 2026 the company was reported to be in deal talks at an approximately $7 billion follow-on valuation.

Affiliations

• Google Brain: Research scientist (Google AI Resident, then full research scientist), 2017 to April 2023.
• Google DeepMind: Senior research scientist and chemistry/physics/materials-science research lead, April 2023 to March 2025.
• Periodic Labs: Co-founder, March 2025 to present.

Notable contributions

Cubuk's body of public work spans data-augmentation methodology for general deep learning, model-robustness and semi-supervised learning, the GNoME materials-discovery research line, and the founding of Periodic Labs.

• AutoAugment: Learning Augmentation Strategies from Data (May 2018, CVPR 2019). Lead author with Barret Zoph, Dandelion Mane, Vijay Vasudevan, and Quoc Le at Google Brain. Introduced reinforcement-learning-based search over data-augmentation policies and produced state-of-the-art image-classification accuracy on CIFAR-10, CIFAR-100, and ImageNet at publication. Combined citation count exceeds 6,000.
• RandAugment: Practical Automated Data Augmentation with a Reduced Search Space (September 2019, NeurIPS 2020). Lead author with Zoph, Jonathon Shlens, and Quoc Le. Simplified AutoAugment into a two-hyperparameter alternative that matched learned-policy performance and became standard practice for vision-model training. Over 5,000 citations.
• SpecAugment (April 2019, INTERSPEECH 2019). Co-author with Daniel Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, and Quoc Le. Extended data augmentation to automatic speech recognition; became a default training step for production speech models.
• Scaling Deep Learning for Materials Discovery (Nature, November 2023). Senior author with Amil Merchant (lead), Simon Batzner, Samuel S. Schoenholz, Muratahan Aykol, and Gowoon Cheon at Google DeepMind. Introduced GNoME and reported 2.2 million predicted stable inorganic crystal structures, with approximately 380,000 estimated stable at zero kelvin. The accompanying autonomous-laboratory paper (Szymanski et al., 2023) demonstrated robotic synthesis of selected GNoME predictions and was the methodological precursor to the Periodic Labs autonomous-laboratory architecture.
• Periodic Labs co-founding (March 2025). Co-founder alongside Liam Fedus. The company has raised $300 million in seed financing at a $1.3 billion valuation, established its San Francisco headquarters, and stated its first commercial target as new superconducting materials.
• Public-talk record. Building an AI Physicist: ChatGPT Co-Creator's Next Venture on the a16z Podcast with Anjney Midha and Liam Fedus (September 2025); DeepMind announcement of GNoME (November 2023); seminar appearances at MIT and Stanford on computational materials discovery.

Investments and boards

The entries below are limited to AI, semiconductors, datacenters, software, and energy.

• Periodic Labs (AI): Co-founder, March 2025 to present. San Francisco-based AI-for-science lab. Cumulative funding $300 million through April 2026 at a $1.3 billion seed-round valuation, with March 2026 reports of deal talks at an approximately $7 billion follow-on valuation.

No other public investor activity on record in AI, semiconductors, datacenters, software, or energy as of May 2026.

Network

Cubuk's longest-running professional collaboration is with Liam Fedus, his Periodic Labs co-founder. The pair met at Google Brain approximately seven years before the Periodic Labs founding and developed the AI-scientist thesis through ongoing conversations during Cubuk's Google DeepMind period and Fedus's OpenAI period.

His Google Brain and Google DeepMind period produced his closest research collaborations. The data-augmentation research line connected him to Barret Zoph, Quoc Le, Jonathon Shlens, and Daniel Park. The GNoME research line connected him to Amil Merchant, Simon Batzner, Samuel Schoenholz, Muratahan Aykol, and Gowoon Cheon, several of whom became Periodic Labs early hires. The autonomous-laboratory research connected him to Nathan Szymanski and the Lawrence Berkeley National Laboratory experimental group. His Harvard PhD under Efthimios Kaxiras anchors his physics-research network, with adjacent ties to the Reed group at Stanford from his postdoctoral period.

Among DeepMind senior leadership, his closest peer relationships include Demis Hassabis, Koray Kavukcuoglu, and John Jumper, alongside Jeff Dean, who participated in the Periodic Labs seed round. Among Insurgent-lab founder peers, his Periodic Labs position runs in parallel with Mira Murati and Barret Zoph at Thinking Machines Lab, Ilya Sutskever at Safe Superintelligence, and Misha Laskin at Reflection AI, with closer thematic adjacency to Isomorphic Labs's drug-discovery framing.

Position in the field

As of May 2026, Cubuk occupies a structurally distinctive position among Insurgent-lab co-founders through the combination of applied-physics doctoral training, a Google Brain to Google DeepMind tenure spanning data-augmentation methodology and materials-science discovery, the GNoME research line as senior author, and the AI-for-physical-sciences thesis at Periodic Labs.

Industry coverage has characterized Cubuk as the materials-science anchor of the Periodic Labs founding pair, complementing Liam Fedus's frontier post-training background. The September 2025 seed round was the largest disclosed seed round in venture-capital history at the time, exceeded subsequently by Mira Murati's Thinking Machines Lab and by Periodic Labs's own reported follow-on talks. His role at Periodic Labs is the direct industrial extension of the closed-loop autonomous-laboratory methodology pioneered in the November 2023 GNoME and Szymanski et al. Nature papers.

The applied-physics-to-frontier-AI arc has parallels in the senior frontier-research cohort. John Jumper, John Schulman, and Misha Laskin share the trajectory of physical-sciences quantitative training preceding a transition into deep learning. Cubuk's record is distinctive in the explicit return to the physical sciences as the commercial application of frontier-AI methodology.

Outlook

Open questions over the next 6 to 18 months:

• First demonstrated superconductor result. Whether Periodic Labs produces a publicly disclosed superconducting-materials discovery from the autonomous-lab platforms, the timing of the first publication, and the magnitude of the capability claim. Cubuk's GNoME background positions him as the principal scientific signatory on any first capability claim.
• Follow-on financing. Whether the reported deal talks at an approximately $7 billion valuation close, the lead investor, and the capital deployment plan for autonomous-laboratory infrastructure scaling.
• Application-area expansion. Whether the company expands beyond the publicly stated superconductor focus to additional materials, energy-storage, semiconductor, or pharmaceutical applications, and the structure of any commercial partnerships in those areas.
• Senior-talent recruitment. Continued movement of materials-science, computational-chemistry, and AI-research staff from Google DeepMind, academic groups, and industrial-research organizations into Periodic Labs's research bench.
• Public-commentary cadence. Frequency and substance of conference appearances and technical publications as Periodic Labs moves from stealth to first commercial results.
• Connection to the broader Google DeepMind materials-research line. Whether the GNoME and AlphaFold-adjacent scientific-AI lines at DeepMind continue independently, or whether technology-transfer arrangements emerge between the published Google research and the Periodic Labs commercial program.

Sources

• Ekin Dogus Cubuk. LinkedIn profile with Periodic Labs, Google DeepMind, Google Brain, Stanford, and Harvard chronology.
• Ekin Dogus Cubuk on X. Public X account.
• Ekin Dogus Cubuk on Google Scholar. Publication record and citation counts.
• Ekin Dogus Cubuk on the Kaxiras Research Group page. Harvard PhD program record under Efthimios Kaxiras, with research focus on disordered solids and battery materials.
• Ekin Dogus Cubuk on the Reed Group page. Stanford postdoctoral record in the Materials Computation and Theory Group.
• AutoAugment: Learning Augmentation Strategies from Data. May 2018, CVPR 2019, with Cubuk as lead author and Zoph, Mane, Vasudevan, and Le as co-authors.
• RandAugment: Practical Automated Data Augmentation with a Reduced Search Space. September 2019, NeurIPS 2020, with Cubuk as lead author and Zoph, Shlens, and Le as co-authors.
• Scaling Deep Learning for Materials Discovery. Nature, November 2023, with Merchant as lead author and Cubuk as senior author. The GNoME paper.
• Millions of new materials discovered with deep learning. November 2023 Google DeepMind announcement of the GNoME results.
• An autonomous laboratory for the accelerated synthesis of inorganic materials. Nature, November 2023, with Szymanski as lead author and Cubuk as listed contributor. The companion paper to GNoME demonstrating robotic synthesis of predicted materials.
• Ekin Dogus Cubuk runs a startup to accelerate physics R&D using AI. Physics Today profile of the Periodic Labs founding thesis with biographical chronology.
• Top OpenAI, Google Brain researchers set off a $300M VC frenzy for their startup Periodic Labs. October 2025 TechCrunch coverage of the seed-round backstory and the founders' Google Brain history.
• Building an AI Physicist: ChatGPT Co-Creator's Next Venture. September 2025 a16z Podcast episode with Anjney Midha on the Periodic Labs thesis.