“Shortening scientific loops accelerates discovery”

Dr. Satrajit Ghosh is a senior research scientist at the McGovern Institute for Brain Research at MIT and an assistant professor at Harvard Medical School. He has helped advance neuroinformatics, open science, and reproducible neuroimaging through both his research and the development of widely used community tools. His work spans machine learning for neuroimaging, the neural mechanisms of speech, and the use of speech features to inform psychiatric diagnosis and treatment. He earned his bachelor’s degree with honors in computer science from the National University of Singapore and his PhD in cognitive and neural systems from Boston University. He has contributed to influential projects including Nipype, fMRIPrep, and NeuroVault. More recently, he has focused on how shared scientific infrastructure can connect domains, modalities, and scales across neuroscience and help address the field’s growing fragmentation.

In this episode, Peter and Satrajit discuss the origins of tools like Nipype and the broader push for reproducible neuroimaging, showing how practical research challenges can inspire infrastructure that benefits the entire field. They also explore functional gradients in the brain and cerebellum, the promise of speech as a scalable biomarker for mental health, and the cautious role AI may play in diagnosis and scientific discovery. A major theme in their conversation is the fragmentation in neuroscience, with knowledge often siloed across methods, scales, and communities. Ghosh argues for a more intelligent scientific infrastructure that connects data, tools, theory, and expertise. He closes with advice to young scientists: experiment often, make mistakes, and learn by discovering where systems fail.

We hope you enjoy this episode!

Chapters

00:00 Introduction to Satra Ghosh and His Work

06:46 The Intersection of Control Theory and Speech

11:18 Satra’s Academic Journey into Neuroscience

20:58 Neuroinformatics and Tool Development

34:42 Individual Differences in Brain Structure

39:21 Developing tools to augment Experimental Design

44:25 Building an Intelligent Infrastructure for Neuroscience

58:45 The Role of Theory in Neuroscience

01:00:26 Access to Scientific Research Expediting Progress

01:06:40 Experience Inherent to Learning

01:09:33 Mapping the Brain’s Functional Gradient

01:16:31 AI and Speech Analysis in Mental Health

01:29:31 Advice, Fail More, Learn More

Works mentioned:

34:59 - Marek, S. et al. (2022). Reproducible brain-wide association studies require thousands of individuals. https://doi.org/10.1038/s41586-022-04492-9

43:44 - Ghosh, Satrajit (2025). An Intelligent Infrastructure as a Foundation for Modern Science.

https://doi.org/10.48550/arXiv.2508.10051

01:09:33 - Margulies, Daniel S., et al. (2016). Situating the default-mode network along a principal gradient of macroscale cortical organization. https://doi.org/10.1073/pnas.1608282113

01:10:13 - Xavier Guell, Jeremy D Schmahmann, John DE Gabrieli, Satrajit S Ghosh (2018). Functional gradients of the cerebellum. https://doi.org/10.7554/eLife.36652

Tools and resources mentioned:

Nipype : an open-source Python framework for building reproducible neuroimaging workflows.

https://nipype.readthedocs.io/en/latest/index.html

fMRIPrep : a robust, analysis-agnostic preprocessing pipeline for functional MRI. https://fmriprep.org/en/stable/

OpenScope : an open-science effort for large-scale neuroscience data sharing and analysis.

https://www.allenneuraldynamics.org/projects/openscope

DANDI : a platform for publishing, sharing, and processing neurophysiology data.

https://about.dandiarchive.org/

NeuroVault : A public repository of unthresholded statistical maps, parcellations, and atlases of the brain.

https://neurovault.org/

Episode producers:Ömer Faruk Gülban, Karthik Sama

Neurosalience #S2E5 with Jack Gallant - Strong opinions about fMRI analysis