The National Institutes of Health has awarded UC San Diego $4.85 million to expand the NeuroElectroMagnetic Data Archive and Tools Resource (NEMAR) into a high-performance computing hub for large-scale neuroelectromagnetic research.
The award (R24-MH120037) funds work from April 2026 through December 2030 and is part of the NIH BRAIN Initiative resource for sharing neuroelectromagnetic data. NEMAR will mount curated electroencephalography (EEG) and magnetoencephalography (MEG) datasets directly on the San Diego Supercomputer Center’s Expanse supercomputer. That arrangement lets researchers run petabyte-scale analyses “in place” without transferring large files to distant compute centers.
NEMAR connects data from OpenNeuro and other sources with the Neuroscience Gateway and SDSC compute. Users will be able to run common science software — EEGLAB, MATLAB, Python and machine learning frameworks such as TensorFlow and PyTorch — against datasets stored on Expanse.
The new funding will also expand NEMAR’s AI work. The team plans to develop multimodal foundation models trained on neuroelectromagnetic signals plus behavioral and participant-level metadata. The models are intended to support tasks such as automated data-quality checks, cross-modal analyses, cognitive-state decoding and brain–computer interface research, the project says.
Project co-principal investigators include Scott Makeig, Amitava Majumdar, Taylor Berg-Kirkpatrick and Arnaud Delorme at UC San Diego, and Russ Poldrack at Stanford. Collaborators named in the award include Srikantan Nagarajan (UC San Francisco) and Kay Robbins (UT San Antonio). SDSC staff Subha Sivagnanam, Choonhan Youn and Yahya Shirazi are listed as contributors.
The platform will continue to enforce data standards to support cross-study machine learning. NEMAR uses the Brain Imaging Data Structure (BIDS) for formatting and the Hierarchical Event Descriptor (HED) system for event annotations. The team also plans workshops and tutorials on data standards, signal processing and machine learning for HPC-enabled neuroscience workflows.
Amitava Majumdar said colocating large datasets with SDSC compute reduces the time researchers spend on data logistics and broadens access to compute-intensive analyses.
Photo credit: today.ucsd.edu
Tags: EEG, MEG, high-performance computing, neuro-AI, BIDS
Topics: EEG & neuro-sensing headsets, Brain–computer interfaces, Neuroscience & neuroplasticity