Vancouver faculty supports sustainability through wildfire management and AI efficiency

The power of machine learning and computer vision now play a key role in wildfire management, in part thanks to research from Ryan Rad, an assistant professor with the Khoury College of Computer Sciences at Northeastern University in Vancouver. His research focuses on using AI for environmental sustainability and making the AI systems themselves more resource efficient.

“There are three things we can do with AI in wildfire management: detect active fires, predict where they might occur, and spread modeling,” Rad said. “Two thirds of wildfires have been caused by humans in the last ten years, and there’s no way to predict a human-caused wildfire. But spread modeling is a way to predict the behavior of an active wildfire using data from the region and satellites, which allows resources to be placed more strategically to handle the fire.”

Spread modeling is a way to predict the behavior of an active wildfire using data from the region and satellites, which allows resources to be placed more strategically to handle the fire.

Rad sees his work translating into real-world impact with the City of Vancouver and other government and community organizations. One of his student groups is conducting research on urban resource management using multimodal AI. Park maintenance requires a lot of labor, and if maintenance is not performed at the right time, the problems grow exponentially.

“We’re building an agentic system for the City of Vancouver to manage their park assets,” he explained. “And on partnerships to provide better disaster predictions and communication. Microsoft’s Vancouver office is working with us. Models are not yet that accurate and alerts are very generic. We’re working on improving knowledge dissemination to ensure people know what to do differently in the moment.”

Meanwhile, Rad is working to optimize large language models for lower latency. He utilizes a hybrid approach involving speculative sampling and knowledge distillation, effectively compressing the model’s knowledge into a smaller, faster architecture without sacrificing performance.

Working both across academia and industry, Rad brings his real-world problem sets to his teaching in the form of case studies and projects. He gives an example from his Generative AI course where students work on a vision language model. In collaboration with Sony, he uses 2D game scenes to build diagnostic framework on rule following, visual QA, and end-to-end gameplay as a testbed to understand the true capabilities and limitations of Vision-Language Models. This research aims to pinpoint exactly where current models fail in order to build better, more robust VLMs.

Photo by Jimmy Jeong

“I always give students challenges that we have a hard time solving in industry,” he explained. “Sometimes I bring case studies that are solved and help them see the path to that solution. A lot of times, students think there is one solution, but when you bring open problems from industry, you give them exposure to a situation where there is no perfect answer. I show how to approach a problem and leave it for others coming after us to solve. From my experience of hiring people in industry, I always teach a course in a way that if I have a position tomorrow, the students are qualified for it.”

Rad sees AI having a huge impact on Vancouver’s tech scene and in industries like healthcare and finance. He believes that Northeastern is playing a key role in Vancouver’s tech-driven future.

One of the reasons I love being at Northeastern is that we are agile and adapt quickly. The entire campus cares about the students.

“One of the reasons I love being at Northeastern is that we are agile and adapt quickly,” Rad said. “The entire campus cares about the students.

Rad highlights that the Northeastern faculty and staff community measures and shares different practices every term to see if they translate to student success in areas like hiring and employer feedback. He especially attributes student success to Northeastern’s encouragement of networking with industry professionals through the regular events programming and through the embedded partner program that fosters employer relationships in the same building.

“We bring everyone to our campus rather than pushing our students to go to places they aren’t comfortable with,” he continued. “Whenever I talk to random people in the city, most people know about Northeastern, even though the campus is less than five years old. People know we exist, and they know the impact we’re having. I hear that immediately.”