Re-purposing Large Language Models for Clinical Prediction
Ahmed Alaa, Professor
Electrical Engineering and Computer Science
Applications for Spring 2024 are closed for this project.
Large language models (LLMs) pre-trained on large corpora of text have
demonstrated incredible capabilities across with zero- or few-shot performance in new tasks that differ from their pre-training objectives. In this projects, we will study the zero- and few-shot performance of LLMs repurposed to issue predictions of clinical outcomes using electronic health record (EHR) data, where longitudinal EHR trajectories are serialized into natural-language strings that are then used to prompt LLMs.
Qualifications: Successful applicants will work with de-identified EHR data to conduct the project above. Successful applicants should have knowledge of Python and strong background/interest in machine learning and NLP. Students are expected to meet with the faculty mentor twice a week, and will be trained to conduct literature review, formulate research problems and engage in scientific writing.
Hours: 12 or more hours
Digital Humanities and Data Science Engineering, Design & Technologies