Interpretable ML for Hidden Systems
Welcome!
I develop machine learning methods to understand opaque processes. My work focuses on creating interpretable models for organizations, processes, and behaviors that are otherwise resistant to measurement. By combining AI, statistics, and domain expertise, I model outcomes and processes in difficult-to-model domains. I have worked on problems in diverse contexts, including predicting mission drift in clandestine organizations, explaining subtexts in international organization negotiations, and identifying potential substitutes in international supply chains for advanced materials.
Curiosity and creativity drive my technical work. I adapt emerging tools while designing products that stay anchored in the big picture. My background as a qualitative researcher gives me a unique sensibility — one that values context, complexity, and collaboration across both technical and non-technical teams.
I am currently working on a project to develop a bespoke Bayesian Item Response Theory model (IRT-M) that solves a longstanding latent variable modeling problem by allowing users to quickly estimate latent dimensions with substantive meaning. Beyond the methods, I’ve turned this into a living tool via an R package and applied it across multiple research projects. This work recieved the inaugural Margaret Levi Award for the Advancement of Comparative Methodology from the Comparative Politics Section of the American Political Science Association.
My work, and work that I have contributed to, appear in the American Political Science Review, Political Science Research and Methods, Studies in Conflict and Terrorism, Third World Quarterly, and International Studies Quarterly. I have consulted for international organizations, including the United Nations Department of Peace Operations, on strategies to use multi-modal data to measure hard, but important, concepts.
I thrive in translation roles where I bridge disciplines and spot opportunities to create new analytic products out of existing models — expanding their reach and amplifying their impact.
I develop machine learning methods to understand opaque processes. My work focuses on creating interpretable models for organizations, processes, and behaviors that are otherwise resistant to measurement. By combining AI, statistics, and domain expertise, I model outcomes and processes in difficult-to-model domains. I have worked on problems in diverse contexts, including predicting mission drift in clandestine organizations, explaining subtexts in international organization negotiations, and identifying potential substitutes in international supply chains for advanced materials.
Curiosity and creativity drive my technical work. I adapt emerging tools while designing products that stay anchored in the big picture. My background as a qualitative researcher gives me a unique sensibility — one that values context, complexity, and collaboration across both technical and non-technical teams.
I am currently working on a project to develop a bespoke Bayesian Item Response Theory model (IRT-M) that solves a longstanding latent variable modeling problem by allowing users to quickly estimate latent dimensions with substantive meaning. Beyond the methods, I’ve turned this into a living tool via an R package and applied it across multiple research projects. This work recieved the inaugural Margaret Levi Award for the Advancement of Comparative Methodology from the Comparative Politics Section of the American Political Science Association.
My work, and work that I have contributed to, appear in the American Political Science Review, Political Science Research and Methods, Studies in Conflict and Terrorism, Third World Quarterly, and International Studies Quarterly. I have consulted for international organizations, including the United Nations Department of Peace Operations, on strategies to use multi-modal data to measure hard, but important, concepts.
I thrive in translation roles where I bridge disciplines and spot opportunities to create new analytic products out of existing models — expanding their reach and amplifying their impact.