My research brings statistical and machine learning methods to clinical data, mainly from emergency medicine, to improve how risk is predicted and how decisions are made at the bedside.
My methodological foundation is dependence modeling with copulas, where I develop tools for tail dependence and feature selection and apply them to clinical risk prediction.
My work is moving increasingly toward two areas: LLMs in healthcare, especially how they can be evaluated for clinical prediction, and causal inference and causal machine learning with electronic health records.
My broader interests also include population health and epidemiology, particularly cancer and mental health.
Across all of this, I aim for applied methodological research that holds up statistically and matters in clinical and public health practice.