Fulbright Visiting ResearcherLudwig Maximilian University of MunichBiostatistics PhD CandidateJohns Hopkins University
I’m on the academic job market for 2026-27. Job market page · CV · Get in touch
Defensible by design.
How do we make data analysis trustworthy enough to act on?
I study how statistical work gets done, by people and more recently by artificial intelligence (AI) agents, and what it takes for analytical conclusions to hold up. Starting from any dataset, my methods take on three questions: when the work is finished, how much evidence the data still hold after repeated reuse, and how to secure the agents that have begun doing that work. Together they make up my dissertation on trustworthy data science, The Well-Made Analysis. I apply them in global health and policy, where my research is cited in guidance from more than 25 countries and the World Health Organization, as well as in biomedical data science and statistics pedagogy.
Three Questions
Completion criteria that tell an analyst, or an AI agent, when an analysis is finished.
A measure of how much confirmatory evidence a public dataset still holds after many studies have reused it.
Monitoring that detects attacks on AI agents from what they do, step by step.
Sources: Google Scholar, Sage Policy Profiles, and CRAN download logs, October 2026.
Students and faculty have also recognized my classroom work, with the department’s Helen Abbey Award for Excellence in Teaching and a student-selected Teaching Assistant Recognition Award from the Bloomberg School. More on teaching and mentoring.