Open Science, Data Sharing, & Digital Philanthropy
Brainhack
Open ScienceReproducible ResearchCognitive NeuroscienceHigh-Performance ComputingPython
I am an active member of the Brainhack Global community, and organized the DC-area brainhacks: https://brainhackdc.github.io/
Gau, R., Noble, S., … Merchant, J. S., … & Zuo, X. N. (2021). Brainhack: developing a culture of open, inclusive, community-driven neuroscience. Neuron, 109(11), 1769–1775.
Supercomputing
Most researchers who need high-performance computing, containerized pipelines, or preregistration never get taught any of it. They pick it up from a labmate, or they do not pick it up at all. A lot of the reproducibility problem is really a training problem.
I have spent a decade trying to close some of that gap:
- Led the Neuroscience and Cognitive Science (NACS) Methods Seminar at the University of Maryland (2019–2022), and authored open training materials on high-performance computing that were adopted across the program — github.com/UMD-COMBINE/IntroToHPCs
- Co-Organized Brainhack Global DC (2018–2021), a workshop series spanning NIH, Georgetown, and the University of Maryland, fostering open-source tool development and hands-on methods training. The broader Brainhack community’s practices were documented in Neuron.
- Co-organized the UMD Epidemiology & Biostatistics seminar series (2022–2025).
- Delivered workshops on fMRIprep and reproducible preprocessing, preregistration in neuroimaging, representational similarity analysis, supercomputing for neuroimaging, and open science principles — at UMD, Georgetown, George Washington University, and the University of Oregon.
- Taught graduate fMRI methods, undergraduate research methods and statistics, and guest lectured on brain networks and cognition at University of Oregon, Georgetown University, and University of Maryland.
Mutual Aid
Civic TechnologyHealth EquityNetwork AnalysisPython
An all-volunteer mutual aid network runs on coordination, and coordination is where volunteer capacity goes to die. Matching offers of help to specific requests, working out who drives where, and getting the right message to the right person every week is a substantial recurring administrative load — the kind that quietly caps how many neighbors an organization can actually reach.
Since March 2020 I have built and maintained the automation that handles it: Google Colab notebooks that match donations to community member needs, plan delivery routes using network science and mapping APIs, and generate the weekly driver-and-recipient assignments. The systems save dozens of volunteer hours per week and let a small volunteer group operate at a scale it could not otherwise sustain.
Between March 2020 and the end of 2024, the organization collected and distributed more than $210,000 in direct aid to over 4,000 individuals and families.
It is also the clearest example I have of what I think good analytics work looks like: unglamorous infrastructure that removes friction so people can do the actual work.