Posted 10:56 a.m. Thursday, May 15, 2025
Graduate student uses AI to tackle animal-vehicle collisions
UW-La Crosse graduate student Kaaviyashri Saraboji is blending her compassion for animals and her drive to design meaningful software.
The UWL software engineering student is developing a new software to help prevent collisions between wildlife and vehicles, a major issue for Wisconsin. In the state, one in every 60 car accidents involved an animal in 2023, making Wisconsin one of the top five states where this type of accident occurs. The artificial intelligence-based system will detect animals near roads before a crash occurs using sensors and machine learning.
“I love animals, and I’ve always wanted to do something meaningful with my work,” she said. “When I heard about how often these crashes happen in Wisconsin, I realized this was a real-world problem that software could help solve.”
The idea took shape during a reflective winter break in January when Saraboji was volunteering at a local nature center. She was thinking about how to apply the practical, hands-on learning emphasized in her UWL courses. Coming from a more theory-focused undergraduate program in India, she admits the transition wasn’t easy.
“My first semester was a struggle,” she says. “Everything here is so applied. The exams, the projects — you’re expected to build things that work in the real world.”
But, with time, Saraboji embraced that challenge. During her break, she began researching animal-vehicle collisions and collecting data and images to train a machine learning model capable of identifying animals in high-risk areas. Over the next several months, she taught herself Python programming and the AI tools needed to bring her idea to life. “This was all new. I had to learn everything,” she said.
Around the same time, Saraboji heard about UWL’s Annual 3 Minute Grad Project competition where she could compete against other graduate students across the university with the challenge of presenting their research clearly and concisely in just three minutes. She decided to enter, meaning she would need to learn even more — perfecting her English pronunciation of words like ‘Montana’ (another high ranking state for animal-vehicle collisions).
Along the way, she found community and support in unexpected places. A coworker and friend at the UWL Card Office helped her practice English pronunciation. Classmates shared insights on real-world problem solving. International students in her residence hall reminded her she wasn’t alone in adjusting to a new culture and academic system.
All of that support — and hard work — paid off in the spring, when Saraboji presented her project in just under three minutes. Despite being new to both public speaking and AI research, she won first place for her clear explanation of how her system could reduce crashes and save lives.
“I immediately texted my friend. I was really surprised,” she said. “They all helped me.”
Saraboji expects to complete the AI model’s training within six months, after which she’ll move on to testing and refining the alert system that will notify drivers in real time.
In addition to her animal detection project, she is also launching her graduate capstone research this summer — development of a secure online application platform for Habitat for Humanity of the Greater La Crosse Region. The system will help families in need apply for affordable housing and home repairs while keeping their personal information safe.
Despite arriving in Wisconsin with little more than a basic knowledge of the Mississippi River, Saraboji now sees the region as a place full of opportunity and community.
“It is a good experience,” she said. “I’ve learned a lot from my friends — and now I can use what I’ve learned to give back.”