Research Program

One Question, Several Threads

My research follows a single question: how do we make quantitative evidence trustworthy enough to act on? I approach it on four fronts at once, from the analytic choices behind a result, to the security of the systems that produce it, to how findings are taught and translated into decisions. The methods work below is the core, and the same throughline runs out to regulatory practice, to the security of the AI tools now doing analysis, and to how the next generation is taught to judge evidence.

Thread 01 · Core Methods

Researcher Degrees of Freedom

A single dataset supports many reasonable analyses. My dissertation builds methods for the choices analysts make: vibration-of-effects and specification analyses that trace how a result depends on those choices. It also asks a question analysts rarely get to answer, “when is an analysis done?”, and treats it as a formal stopping problem. This work is joint with Brian Caffo, Stephanie Hicks, Roger Peng, and Martin Lindquist.

in press
The End is Insight: How to Know When Data Analysis is Done
Columbus, Caffo, Hicks, Peng · under review
first author
2026
Managing Analytic Multiplicity in Epidemiology: Reproducible Multiverse and Vibration of Effects Analyses in R
useR! 2026, the R User Conference, Warsaw
software
An R package for specifying, running, and reporting vibration-of-effects and specification analyses, with a companion Quarto book and source on GitHub.
author and maintainer
Thread 02 · Applied Methods

Trustworthy & Reproducible Data Science for Health and Policy

Methods only matter if their uncertainty travels intact to the people who act on them. As a CERSI Scholar with the U.S. Food and Drug Administration and Johns Hopkins, I study how to report variability from multiple analytic choices in a way that supports regulatory decisions, and how to communicate analytic uncertainty to non-specialist audiences without losing rigor. This thread connects to my applied collaborations on the Global Burden of Disease Study and the National COVID Cohort Collaborative (N3C).

in review
A Multi-Modal Pipeline for COVID-Associated Acute Kidney Injury Drug Repurposing
Columbus et al. · under review at Nature Biotechnology
first author
2025
Strategies to Address and Report Variability Arising from Multiple Analytic Choices in Biomedical Data Analyses
Yale–Mayo Clinic CERSI Scholar Seminar
Thread 03 · AI Safety

Security & Governance of AI for Health and Science

The analysis pipeline is increasingly run by AI systems, which introduces a new class of failures. Drawing on prior work as an information security analyst and data governance specialist, I study how to secure autonomously acting language models and the open-source dependencies they rely on, from prompt injection to data poisoning. I advise on AI safety and governance and have written three O’Reilly Media chapters across application and information security; the full list is on the publications page.

2025
From Traditional to Agentic AI: Securing Autonomously Acting Language Models Against Emergent Threats
Paper, EurIPS 2025 Foundations of Language Model Security Workshop
solo author
Thread 04 · Education

Statistics & Data Science Education

Teaching people to weigh evidence is the same problem as my methods work, viewed from the classroom. I design instruction and assessment for analytic judgment that hold up when students have AI assistants, work recognized by an American Statistical Association (ASA) Best Contributed Paper Award and a Johns Hopkins Teaching Academy Fellowship.

2026
When is an Analysis “Done”? Teaching Analytic Judgment and AI-Aware Assessment in Data-Driven Courses
European University Alliance for Global Health Conference on Pedagogical Development, University of Alcalá
2026
Discussion of “Addressing the Challenges of AI-Generated Assignment Submissions in Education”
Journal of Data Science
solo author
Funding

Grants & Fellowships

My current work is supported by a Fulbright research award, which funds my program at the Ludwig Maximilian University of Munich on the analytic choices behind robust results. Earlier, I held an open-source software grant from the R Consortium, part of the Linux Foundation, and undergraduate research fellowships at the University of California, Irvine. On larger collaborative projects, I’ve contributed as named personnel to federally funded research with the National Aeronautics and Space Administration and the National Science Foundation.

Appointments

Where I Hold Roles

Beyond Munich and Johns Hopkins, I hold research roles that keep me close to large-scale applied work. I’m a Senior Biostatistician and Collaborator with the Institute for Health Metrics and Evaluation, home of the Global Burden of Disease Study, and I was a Team Science Fellow and Visiting Scholar at Stanford Medicine. I’m also a Fellow of the Open Science Center at the Ludwig Maximilian University of Munich and a Scholar in the Center of Excellence in Regulatory Science and Innovation with the U.S. Food and Drug Administration and Johns Hopkins.

Full publication list and metrics on the publications page, or browse Google Scholar.

Honors & Awards

Selected Recognition

A few of the honors I’ve been grateful to receive, across research, teaching, and early-career work.

2026
Helen Abbey Award for Excellence in Teaching
Johns Hopkins Department of Biostatistics
2026
Trailblazers in Engineering Fellow
Purdue University College of Engineering
2025
Fulbright Research Award
Fulbright U.S. Student Program · with the Germanistic Society of America Quadrille Ball Fellowship
2025
Best Contributed Paper Award
American Statistical Association, Statistics and Data Science Education Section
2025
Jane and Steve Dykacz Award
Best Student Paper in Medical Statistics, Johns Hopkins Department of Biostatistics
2025
ISyE-MS&E-IOE Rising Star
Georgia Tech, Stanford, and the University of Michigan
2022
Vivien Thomas Scholars Initiative PhD Fellowship
Johns Hopkins Office of the Provost · first Vivien Thomas Scholar in the Bloomberg School of Public Health
2018
Phi Beta Kappa & Chancellor’s Award of Distinction
University of California, Irvine
2018
NASA Datanaut
one of fifty worldwide selected for the Spring 2018 class