Stanford
University
  • Stanford Home
  • Maps & Directions
  • Search Stanford
  • Emergency Info
  • Terms of Use
  • Privacy
  • Copyright
  • Trademarks
  • Non-Discrimination
  • Accessibility
© Stanford University.  Stanford, California 94305.
researchGrant

HAI and Wu Tsai Neuro Partnership Grant

Status
Closed
Topics
Sciences (Social, Health, Biological, Physical)
Overview
2026 Grant Recipients
2022 Grant Recipients
Overview
2026 Grant Recipients
2022 Grant Recipients
Share
Link copied to clipboard!

2026 Grant Recipients

HAI and Wu Tsai Neuro Partnership Grant | Stanford HAI

Major depressive disorder affects hundreds of millions of people worldwide, yet patients often respond very differently to the same treatments. One reason is that depression is not a single biological condition but a collection of distinct brain-based subtypes. Recent research using brain imaging has shown that patterns of brain network connectivity measured with functional magnetic resonance imaging (fMRI) or electroencephalography (EEG) can identify biologically meaningful "biotypes" of depression that predict which treatments are most likely to work. Despite this promise, EEG-based subtyping and fMRI-based biotyping have developed independently, leaving a critical unanswered question: does EEG reflect the same underlying brain circuit biotypes identified with fMRI, or does it capture different forms of biological heterogeneity? This project will address that gap by developing an artificial intelligence framework that directly links EEG and fMRI. Using a uniquely rich dataset collected at Stanford, including paired EEG, MRI, fMRI, and clinical treatment outcomes, we will test whether EEG can reliably replicate validated fMRI depression biotypes at the individual level. We will also explore whether combining EEG and fMRI into a single multimodal model yields more stable and clinically useful biotypes than either modality alone. Finally, we will assess how well these biotypes predict response to brain stimulation treatment. Every other field of medicine has tools like this; it is time psychiatry does too. By integrating neuroscience and human-centered AI, this work aims to lay the foundation for scalable, biologically grounded tools that move depression care beyond trial-and-error and toward truly precision psychiatry.

Name

Role

School

Department

Cammie Rolle

Main PI

School of Medicine

Psychiatry and Behavioral Sciences

Ellen Kuhl

Co-PI

School of Engineering

Mechanical Engineering

Talk therapy helps millions of people with depression and anxiety, but we know almost nothing about what happens in the brain during a therapy session. This gap prevents us from understanding why therapy works for some people and not others, and from developing more personalized treatments. Our project will create the first detailed map of brain activity during real psychotherapy sessions. We will work with patients who have electrodes temporarily implanted in their brains for epilepsy treatment but with comorbid depression, with special interest in electrodes that reach brain regions involved in mood and emotion. While these patients receive therapy sessions during their hospital stay, we will record their brain activity with millisecond precision. To make sense of what we record, we will use artificial intelligence in two ways. First, we will use modern language AI to automatically analyze the therapeutic conversation, identifying when different techniques are being used and what symptoms are being discussed. Second, we will use a mathematical approach called topological data analysis to find latent patterns in how brain activity changes throughout therapy, patterns that would be invisible to conventional analysis. By linking conversation to brain activity moment by moment, we aim to discover which therapeutic techniques engage which brain circuits, how the brain's state changes over a session, and whether we can predict from early brain patterns who will improve. This foundational work will guide the development of more targeted, personalized mental health treatments and improved psychoeducation.

Name

Role

School

Department

Corey Keller

Main PI

School of Medicine

Psychiatry and Behavioral Sciences

Manish Saggar

Co-PI

School of Medicine

Psychiatry and Behavioral Sciences

The human brain contains thousands of distinct types of neurons which are the foundation of all basic and higher cognitive brain function. During development, cells in the brain reliably adopt the correct identity, yet we still lack a clear "rulebook" for how those choices are made. Evidence suggests that the genome encodes this rulebook as instructions read by combinations of DNA binding regulatory proteins that modulate specific gene expression programs giving rise to distinct cell fate. Our project aims to identify the combinations of regulatory proteins that are sufficient to drive a cell into one neuron type rather than another. We will answer this by systematically turning selected regulators on or off in human cells and measuring how gene activity and the cell's regulatory landscape change in each cell and over time. We will then train and interpret AI models on these perturbation datasets together with large public brain cell atlases to learn predictive rules that connect regulatory "instructions" to resulting cell identity. The outcome will be a practical, interpretable map from inputs (regulatory protein combinations) to outputs (neuron types and trajectories), including model-generated hypotheses for minimal subsets of proteins needed to produce specific neuronal cell types. We will validate top predictions experimentally by attempting to generate several clinically and biologically important neuron types that remain difficult to produce reliably in the lab. By revealing how neuron identities are specified, this work will provide unprecedented access to the cellular building blocks of the human brain, an important milestone towards accelerating cell-based therapies in neurological diseases.

Name

Role

School

Department

Marius Wernig

Main PI

School of Medicine

Pathology

Anshul Kundaje

Co-PI

School of Medicine; School of Engineering

Genetics; Computer Science

Have you ever smelled something (fresh-baked cookies, rain on pavement, a specific perfume) and been instantly transported to a specific memory, complete with emotions? Scent has an almost magical ability to unlock our personal histories in ways that other senses don't.

Scientists have recently built AI systems that can predict how a molecule will smell just from its chemical structure, an impressive feat. However, these systems only capture the average response across many individuals. They miss something crucial: what a smell means to each individual. Your emotional response to vanilla might be completely different from someone else's, shaped by your unique life experiences and memories.

We're proposing to peek inside the human brain while people smell different odors and describe what comes to mind, whether that's a simple word like "sweet" or a rich memory like "my mother's kitchen during the holidays." Using rare brain recordings from epilepsy patients, we can watch in real time as smell, emotion, and memory circuits work together. By combining these brain signals with AI language models that understand the personal stories people share, we aim to discover how the brain links sensation to meaning, and whether we can teach AI systems to do the same.

This work could allow us to build smarter, more human-centered AI that doesn't just recognize patterns, but actually understands context, emotion, and what things mean to individual people.

Name

Role

School

Department

Todd Coleman

Main PI

School of Engineering; School of Medicine

Bioengineering

Emily Fox

Co-PI

School of Humanities and Sciences; School of Engineering

Statistics; Computer Science

Human infants are uniquely smart, learning about their visual world more efficiently than many Artificial Intelligence (AI) models available today. Infant learning is even more miraculous when you consider the impoverished visual information they learn from. For instance, infants typically don't interact with a random sample of people, but instead, the bulk of their interactions are with a small circle of family, friends, and caregivers. How do these biased experiences influence the visual expertise of infants, and how do infants generalize beyond these experiences to recognize new people? One way to make progress on this question while also advancing AI research is to build an AI model of the infant mind using the experiences and brain data of an infant. We recently collected a large and holistic longitudinal dataset, which includes recordings of the experiences of a single infant, the brain responses of that infant, and the visual expertise of that infant. We will train an AI model using the experiences of the infant and guide the model training using the infant's brain activity. If we are successful, the model will be a tool for exploring infant development and asking questions that cannot be asked practically (e.g., what happens if the model does not ever see a specific person in its training data). The model may also inspire approaches to build AI models with less data and thus lower their environmental and ethical footprint.

Name

Role

School

Department

Cameron Ellis

Main PI

School of Humanities and Sciences

Psychology

Michael Frank

Co-PI

School of Humanities and Sciences

Psychology

Dan Yamins

Co-PI

School of Humanities and Sciences; School of Engineering

Psychology; Computer Science

Skip to content
  • About

    • About
    • People
    • Get Involved with HAI
    • Support HAI
    • Subscribe to Email
  • Research

    • Research
    • Research Programs
    • Grants
    • Marlowe (opens in new tab)
    • Student Affinity Groups
    • Centers & Labs
    • Research Publications
    • Research Partners
  • Education

    • Education
    • Executive and Professional Education
    • Government and Policymakers
    • K-12
    • Stanford Students
  • Policy

    • Policy
    • Policy Publications
    • Policymaker Education
    • Student Opportunities
  • AI Index

    • AI Index
    • AI Index Report
    • Global Vibrancy Tool
    • People
  • News
  • Events
  • Industry
  • Centers & Labs
Navigate
  • About
  • Events
  • AI Glossary
  • Careers
  • Search
Participate
  • Get Involved
  • Support HAI
  • Contact Us

Stay Up To Date

Get the latest news, advances in research, policy work, and education program updates from HAI in your inbox weekly.

Sign Up For Latest News