Yonghui Wu, Ph.D., stands beside HiPerGator, which supported his trailblazing AI research. Photo courtesy: J. Harper/University of Florida
As anyone who’s filled out forms at a doctor’s office knows, health care providers collect massive amounts of data. It is a time-consuming and labor-intensive process for doctors and staff to record that information and retrieve and analyze it quickly when needed.
HiPerGator has a fix for that. Enter GatorTron™.
Developed by researchers from UF Health and NVIDIA in 2021, GatorTron is the largest clinical language model in the world. (An AI model is a program that applies algorithms to data to recognize patterns, make predictions or make decisions without human input.) It was trained on 90 billion words taken from the electronic health records of more than 1 million patients in the UF Health system, anonymized to preserve individuals’ privacy.
GatorTron’s superpower is its ability to process notes taken by health care workers and generate summaries, inferences and answers with a high degree of accuracy (more than 89%, according to a recent study). This sets the stage for multiple medical applications, each developed to perform specific health care tasks.
To date, the free versions of GatorTron models have more than 430,000 downloads from Hugging Face, an open-source AI website, and have generated multiple spin-off applications. 
WHY IT MATTERS
UF’s GatorTron is the most robust AI health-care language tool in the world