By Ines Aviles Spadoni, M.S., M.A., Research & Communications Coordinator
Helen Michael, a rising senior at the University of Texas (UT) at Austin, is at UF this summer as part of the Secure, Accessible, and Sustainable Transportation Research Experiences for Undergraduates (SAST REU) – a program funded by the National Science Foundation (NSF) and supervised by faculty affiliated with the Department of Electrical and Computer Engineering, the Warren B. Nelms Institute for Connected World, and the University of Florida Transportation Institute (UFTI). Michael’s curiosity in data science and making a real-world impact drew her to pursue the internship.
“I was interested in this opportunity because I’m passionate about exploring the various applications of data science, especially in a domain that is impactful on a large scale like transportation,” she said. “I saw this as a chance to apply my technical background to a field I hadn’t worked in before, while also contributing to research that could help make roads safer and transportation systems more sustainable.”
Michael’s summer is spent working with Siva Srinivasan, Ph.D., professor of civil engineering and associate director of the UFTI. Under Srinivasan’s supervision, she is working on fine-tuning large language models (LLMs) to analyze crash reports and police narratives to assess the safety of autonomous vehicle deployment. Srinivasan reflects on how Michael’s unique skills can help to advance this type of research area, including the importance of the SAST-REU program.
“It’s wonderful to work with Helen because she brings a whole new toolbox of skills that we can apply to transportation safety problems,” Srinivasan said. “The NSF-funded site SAST-REU program is a wonderful opportunity for undergraduate students to get involved in research and learn how to apply their knowledge to addressing real-world problems.”
Traditional safety analysis predominantly relies on applying statistical methods to the tabular version of the data from police reports. The crash narratives written by police officers have largely not been analyzed at scale, arguably because of methodological limitations. The LLMs that have emerged in the last couple of years can help overcome this issue.
“It’s a great opportunity to apply data science, especially natural language processing and machine learning to real-world text data and explore how AI can support safer decision-making in autonomous vehicle deployment,” Michael said.
Michael is not majoring in civil engineering and has never taken a transportation engineering class; she’s a mathematics major with an interest in data science. However, she believes that the analytical and technical skills that she is learning in her major can translate across disciplines such as transportation.
At UT Austin, she’s worked on many data science projects that have provided her with skills useful for the transportation industry, such as data cleaning, modeling, and working with large or unstructured datasets.
“Since I’ve applied these skills across different domains, it helps me transition into an application in transportation, and I’m excited for what this project has to offer,” she said. “I’m also eager to apply what I’ve learned in my coursework so far to a real-world project that has social impact.”
Although Michael has never taken a transportation class, she says she is passionate about making sure that she can use her skills to reduce car crashes and traffic-related injuries – an issue that affects millions of people globally and is preventable. An insight that Srinivasan shared with her at the beginning of the internship has impacted her perspective on transportation safety.
“Something Dr. Siva said early on really stuck with me is that we shouldn’t call them ‘accidents’ because that implies that they’re unavoidable, when in fact many can be prevented,” she said. “That perspective has shaped how I think about this work and motivates me to use data-driven research and real-world insights to help improve road safety.”