Education
FSU data scientist receives NSF CAREER Award to develop AI model to analyze complex numerical data at scale
A Florida State University artificial intelligence and data scientist has earned one of the most prestigious early career faculty awards to develop a new type of AI model that better calculates and uses numeric data, including large-scale healthcare data for applications in cancer research.
Michael Gubanov, associate professor in the Department of Computer Science, received a 2026 CAREER Award from the National Science Foundation’s Faculty Early Career Development Program for his work developing an advanced large language model, or LLM, to analyze properties of numeric data, such as number order, and more accurately calculate and interpret numeric data.
The Faculty Early Career Development Program offers the NSF’s most significant awards in support of early career faculty who have the potential to serve as leaders in research and education. The award provides five years of funding to support scientific goals, student training and educational outreach activities.
“While AI can’t diagnose or treat patients, this advanced AI model is expected to help healthcare professionals make more accurate decisions,” Gubanov said. “Because of its ability to better comprehend complex numerical data, it can be applied to gauge the promise of candidate cancer treatments for patients based on individual factors, such as prior cancer treatments, to increase cancer patients’ survival rate.”
Gubanov is collaborating with Moffitt Cancer Center in Tampa, Florida, to apply and evaluate the model’s potential impact in the field of healthcare to further understand how it can be used for personalized recommendations of cancer treatments like chemotherapy and immunotherapy. At FSU, Gubanov is the founder and supervisor of BigLab!, which bridges AI and data science technologies with cancer research to improve the quality of cancer care and treatment.
“Michael has successfully used AI techniques in novel ways to help interpret large-scale health data,” said Sam Huckaba, dean of the College of Arts and Sciences. “There’s a level of complexity that must be overcome to provide swift and meaningful analysis of these rich datasets. Michael is among a handful of researchers who’ve made progress. His NSF CAREER Award is a mark of excellence that highlights his cutting-edge research.”
LLMs determine answers to numeric questions by relying on complex statistical pattern recognition and common themes instead of recognizing numeric qualities or possessing the ability to calculate — generally, models comprehend and encode numbers as tokens. Because of this, LLMs often provide inaccurate responses to numerical operations because the model is largely guessing the most probable answer rather than calculating the answer.
“Models don’t fully comprehend the order and magnitude of numeric data,” Gubanov said. “If you ask an LLM what two plus two is, it will likely say four — not because it calculated what two plus two equals, but because the model recognizes four as the most likely answer to that question. Our goal is to develop a model that can instead apply more meaningful quantitative reasoning to better comprehend numerical data.”
A heightened validity of numerical data analysis can help make AI applications in clinical practice more trustworthy. This new model also has potential applications in industries that require precise calculations of numeric data, marking significant shifts in medicine, government, finance, science and more.
“Currently, LLMs hallucinate and are inherently non-deterministic, which makes them hit accuracy barriers preventing their adoption in medicine and healthcare because they lack high numeric precision,” Gubanov said. “We’re planning to build a model that doesn’t use guesswork and instead understands the properties of numerical data, which is often more complex and structured than natural language text.”
This project will also support the development of new teaching materials, student research opportunities in data science and AI at the BigLab! and outreach activities to engage students in data reasoning. In addition to NSF funding, BigLab! has received support from the Casey DeSantis Cancer Research Programs’ Florida Cancer Innovation Fund and Amazon Web Services.
“Receiving the NSF CAREER Award is a huge honor, but nobody succeeds by themselves,” Gubanov said. “I’m grateful for the inspiration and support from my collaborators and colleagues, FSU, and the BigLab!. Without them all, this project wouldn’t be possible.”
Gubanov earned his doctorate in computer science from the University of Washington in 2010 and completed postdoctoral research at the Massachusetts Institute of Technology before taking an endowed faculty position at the University of Texas at San Antonio. He joined FSU’s faculty in the Department of Computer Science in 2018.
Visit the Department of Computer Science website to learn more about research conducted in the department.
