Kidney transplantation is one of the most effective treatments for kidney failure, yet deceased-donor kidneys remain scarce and many usable organs are ultimately discarded. Current allocation systems offer kidneys sequentially to compatible candidates, but repeated declines and logistical delays can increase cold ischemia time, reducing the likelihood of transplantation. Furthermore, predictive models for identifying kidneys at risk of delayed acceptance have also been developed using data from a limited number of geographic regions and may not account for emerging preservation technologies or differences in patient and clinician preferences. This SuperUROP project will develop and evaluate an uncertainty-aware decision-support system for kidney allocation. The project improves on a clinician-facing interface for communicating model predictions and uncertainty, evaluates model generalizability and failure modes across geographic and clinical subgroups, investigates how emerging preservation technologies can be incorporated into the prediction framework, and gathers patient and provider feedback to guide model and interface design. By combining machine learning, software engineering, and stakeholder-informed design, this project aims to create a more transparent, generalizable, and clinically usable tool for identifying kidneys that may benefit from expedited placement.

I’m very interested in applying machine learning and computational tools to medicine. My goal is to become a clinician-scientist who uses insights from clinical practice to guide the development and implementation of these technologies, while also leveraging them to improve patient care. I’m excited about this interdisciplinary SuperUROP that explores this space!