Tal Zeevi

Tal Zeevi, PhD is a machine learning researcher at Yale University, affiliated with the Image Processing and Analysis Group and the Yale Interventional Oncology Research Lab. His research focuses on predictive uncertainty in deep learning, particularly in medical image analysis where limited data and model underspecification can support multiple clinically plausible interpretations.

His work develops post-training, inference-time methods that characterize how uncertainty is structured and expressed within a trained model, and how this structure affects the delineation of anatomical boundaries and disease extent. A central emphasis of his research is decision stability. He studies whether model-derived clinical conclusions remain consistent under repeated evaluation.

He applies these methods across cancer and neuroimaging applications, including prostate cancer imaging, hepatocellular carcinoma risk modeling, and interventional radiology workflows. His work has received editorial commentary in Radiology, The New England Journal of Medicine, and The American Journal of Roentgenology.

Prior to Yale, he was a research scientist at Intel’s Artificial Intelligence Solutions Group, where he applied machine learning to problems in processor architecture and system design. He holds graduate degrees in engineering and applied statistics, with earlier training in multi-agent systems and game-theoretic modeling.

At Yale, he contributes to graduate and undergraduate teaching in biomedical image processing and research ethics, and holds the Certificate of College Teaching Preparation (2025).