The university will combine MRI scans, wearable brain recordings and Marlowe computing, with datasets and models to be made freely available to researchers worldwide
Stanford University is bringing faculty from its Humanities and Sciences, Medicine and Engineering schools together for a five-year effort to build AI models that can predict how the human brain responds across different activities and conditions.
The Stanford Digital Brain Project, led by the Wu Tsai Neurosciences Institute, will collect brain-imaging data specifically for training foundation models. Researchers aim to use those models to test ideas about brain function in a virtual environment and investigate changes associated with brain disorders.
The project has received lead support from philanthropists Mark and Mary Stevens, with additional support from Clara Wu Tsai.
Its research program connects Stanford’s investments in imaging equipment, shared laboratories and AI computing. Model development will draw on Marlowe, the GPU computing cluster identified by the Stanford Institute for Human-Centered Artificial Intelligence (HAI) as supporting some of the project’s AI training.
Dan Yamins, Associate Professor of Psychology and Computer Science and a Faculty Scholar at the Wu Tsai Neurosciences Institute, has played a leading role in developing the project.
“By training foundation models on brain data at scale, we’ll be able to form quantitative predictions of how brain circuits process information – tools that can help us explore hypotheses about brain function in ways that would be impossible without a digital model,” Yamins says.
MRI and wearable recordings capture where and when the brain responds
The project’s experimental campaigns will examine brain activity underlying perception, memory, language, movement, decision-making, emotion, social interaction and reasoning.
Researchers will collect complementary measurements using a 7-tesla magnetic resonance imaging (MRI) scanner at the Lucas Center for Imaging at Stanford Medicine and a wearable OPM-MEG brain-imaging system at the Koret Human Neurosciences Community Laboratory.
The MRI scanner, unveiled in 2025, provides detailed information about where activity occurs. The wearable system captures its timing with millisecond precision.
Laura Gwilliams, Assistant Professor of Psychology and a Faculty Scholar at Wu Tsai Neuro and Stanford HAI, leads the project’s OPM experiments. She describes experiments in which participants view, describe and recollect thousands of naturalistic videos, allowing researchers to examine how brain networks coordinate perception, attention and memory.
The measurements will form part of a unified experimental framework designed to capture how different brain networks work together during everyday activities.
Kalanit Grill-Spector, the Susan S. and William H. Hindle Professor in the School of Humanities and Sciences and a Professor of Psychology, is helping lead data collection.
Grill-Spector says: “Each new domain of mental function we measure will contribute to a growing body of brain and behavioral data capable of grounding a comprehensive digital model of the human brain.”
Marlowe supports a staged approach to brain modeling
The model development team will use computational methods developed by Yamins and colleagues to train foundation models on the combined datasets.
These models are intended to learn patterns across varied brain-imaging data and generate predictions about responses in circumstances beyond those directly measured during training.
The team will begin with models of individual brain systems, including perception, memory and attention, before expanding toward models of the entire human brain.
Data collection and model development will inform each other throughout the project. Researchers will compare predictions with measured brain activity and behavior, using those checks to assess accuracy and guide further work.
The program will also develop prototype applications for understanding and intervening in brain disease. Those applications remain research goals, with model accuracy and proposed interventions requiring testing.
Clinical research goals and open access for other scientists
Researchers hope the models will help identify subtle changes in brain networks associated with disorders such as depression and Parkinson’s disease.
Potential applications include finding early biomarkers and developing more targeted interventions that could subsequently be examined in clinical studies. The university also describes the possibility of using models to investigate why people respond differently to treatments.
Kang Shen, Director of the Wu Tsai Neurosciences Institute, points to longer-term possibilities such as simulating a proposed brain-circuit intervention before testing it in a patient, or running virtual experiments to develop theories about networks involved in autism or Alzheimer’s disease.
The project is designed to give researchers a way to explore those questions computationally, with predictions checked against brain measurements and human behavior.
Stanford plans to make the resulting datasets and models freely available to researchers worldwide. Its data-sharing approach will build on existing Stanford-developed resources, including the OpenNeuro platform and the Brain Imaging Data Structure used to organize brain-imaging data.
ETIH Innovation Awards 2027 open for entries
The ETIH Innovation Awards 2027 are now open for entries, with 26 categories recognising companies, institutions, projects and individuals making a measurable impact across education technology. The global awards span AI in education, digital learning, higher education, skills and workforce, inclusion, education operations, evidence and impact, with entries managed through ETIH’s new online awards platform.
Entries are open worldwide, with the early-bird deadline on Friday 18 December 2026 and the final entry deadline on Friday 26 February 2027 at 23:59 UK time.
Explore the 2027 categories and judging criteria, read the How to Enter guide or start an entry through the ETIH Innovation Awards 2027 platform.
