I am a machine learning researcher at Yale University, based in the Image Processing and Analysis Group and an affiliated researcher with the Yale Interventional Oncology Research Lab. I study predictive uncertainty in underspecified settings, where multiple competing model interpretations are statistically supported.

My work develops post-training, inference-time methods that map the structure of a model's uncertainty: what alternatives the model considers plausible, how those alternatives relate, and how they are organized within the model's latent representations. A central focus of my research is decision stability, whether a model's predicted decision remains reproducible under the stochasticity induced by underspecification. I frame inference as controlled exploration of the model's hypothesis space, with uncertainty treated as an operational object for decision-making and evaluation, especially in safety-critical domains such as medical imaging.

I spent several years in industry research at Intel AI and led translational machine learning research at Yale through the Translational Image Analysis and Machine Learning (TIAML) Center on core computer-vision problems in medical imaging, with primary applications in interventional radiology and neuroimaging. I also contribute to graduate education in biomedical image processing and analysis and research ethics.

Outside Research, I am trained in music, particularly in arrangement and composition. Music was, in many ways, my earliest mentor for research. It taught me how to search without knowing the outcome, and how multiple interpretations can coexist within the same underlying structure.