Matthew P. Gunn, PhD
Computational neuroscientist and systems engineer with specializing in neuroimaging (EEG/fMRI), multimodal machine learning, and AI-driven analytics.
Matthew has led the design of reproducible data pipelines, autonomous software systems, and explainable AI frameworks across academic, startup, and applied research settings. His work integrates large-scale neural, behavioral, and genetic data to model latent brain states, reward processing, and psychopathology, with a focus on translation to precision mental health and neurotechnology.
He has served as a Postdoctoral Associate at the University of Illinois Chicago and has authored peer-reviewed publications, open-source tools used by 700+ researchers worldwide, and technical frameworks for reliable AI-assisted development.
Selected Work & Scholarship
Selected Publications
Gunn, M. P., Rose, G. M., Whitton, A. E., Pizzagalli, D. A., & Gilbert, D. G. (2024). Smoking progression and nicotine-enhanced reward sensitivity predicted by resting-state functional connectivity in salience and executive control networks. Nicotine & Tobacco Research, 26(10), 1305–1312. https://doi.org/10.1093/ntr/nta123
Anderson, Z., Gunn, M., Jones, E., Ajilore, O., Phan, K. L., de Wit, H., Klumpp, H., Calhoun, V., & Crane, N. A. (2025). Δ9-Tetrahydrocannabinol Alters Limbic and Frontal Functional Brain Connectomes Among Young Adult Cannabis Users. Biological psychiatry. Cognitive neuroscience and neuroimaging, S2451-9022(25)00282-4. Advance online publication. https://doi.org/10.1016/j.bpsc.2025.09.005
Open-Source Systems
NeuroLode (EEGLAB plugin v1.7.0) — Export to Excel/DAT/TXT/ASC; spectral features (centroid, kurtosis, skewness, spread) and modified BSS UI
eeg_preprocess —MATLAB + EEGLAB preprocessing template for EEG datasets. Automates channel selection, filtering, ASR sweeps, ICA, ICLabel rejection, and QC logging.
Spatiotemporal PCA (Data-Driven EEG Decomposition) — MATLAB + EEGLAB pipeline for temporal–spatial factorization of ERP data with parallel analysis, rotation options, and reproducible logging
NeuroToolbox — Generalizable Python package for data-driven neuroimaging analysis. Implements asymmetry indices, residual- and GPR-based normative modeling, multivariate PLS/CCA, network metrics, and robust regressions. Designed for reproducibility, with CLI/YAML workflows and CSV spreadsheet input for easy integration across labs and datasets
Technical Writing
The GPT Engineering Loop (2025): A Field Guide to Reliable AI Coding Workflows (Mini-Book). ISBN: 979-8277384671.
Full CV available upon request.
Rigor before results.
We prioritize robustness, interpretability, and generalization over impressive but fragile outcomes.
Operating Principles
Systems over scripts.
We build reusable, auditable systems—not one-off analyses.
Inference over correlation.
Patterns are not explanations. We emphasize principled inference and causal sensitivity over surface-level associations.
Discretion and restraint.
Not all work is public-facing. Platforms, metrics, and partnerships are disclosed selectively to preserve integrity, privacy, and long-term optionality.
Reproducibility as a requirement.
Outputs must be traceable, repeatable, and reviewable. Reproducibility is an engineering constraint, not an afterthought.
Long-term alignment.
We evaluate decisions not only by immediate performance, but by how well they support durable systems, future extensions, and downstream trust.