The following story originally appeared on the website for the School of Computing, Data Sciences & Physics. – Ed.
A new wave of artificial intelligence aims to help humans make decisions in unpredictable real-world scenarios where digital and physical systems intersect.
From disaster response to public health crises, these systems analyze complex, rapidly changing data and provide actionable insights, guiding leaders and responders to make informed choices under pressure, even when outcomes are uncertain.
Ayan Mukhopadhyay, assistant professor of computer science at William & Mary, has been involved in this field for over a decade, beginning his endeavors during his Ph.D. at Vanderbilt University. Over the years he has collaborated with government, non-profit and for-profit agencies around the world to develop novel AI techniques that aid in solving real-world problems.
One of his latest projects is part of a $1.25 million National Science Foundation grant to participate in the Smart and Connected Communities program (S&CC), which leverages AI to tackle challenges faced by local communities. Mukhopadhyay is collaborating with George Mason University, Old Dominion University and Vanderbilt University on the S&CC project.
The goal of their project is to help the City of Virginia Beach during floods.
But how does AI come into play?
“During these flooding events, it is hard to know which shelters should open, and how to evacuate people effectively. It is a complex decision-making process that depends on many unpredictable factors,” Mukhopadhyay explains. “From a computational standpoint, this is a very high-dimensional optimization problem under uncertainty.”
Their solution is twofold.
First, the goal is to learn data-driven models that predict how interconnected infrastructure might fail. It analyzes years of data from past flooding events — including evacuation outcomes, which shelters were opened or closed, and infrastructure failures such as power outages or unexpected road closures — to forecast potential outcomes in future floods. These predictions are then incorporated into an AI-driven decision-making model that helps officials develop optimal flood-response strategies.
He is also leading a separate Civic Innovation Challenge project with the Nashville Department of Transportation (NDOT) and Vanderbilt University as part a $697,000 NSF grant.
The goal is to help NDOT enhance road safety by automatically identifying illegal road closures.
Nashville is one of the fastest-growing cities in the U.S., and with that growth comes construction and new infrastructure. NDOT issues roughly 40,000 road-closure permits each year and estimates that nearly the same number of closures occur without proper authorization.
“This is not someone stopping to get in an Uber, this is someone doing construction without filing a permit, meaning traffic and local businesses are affected,” explained Mukhopadhyay. “These people are not following official safety precautions either, which puts the community at risk. Plus, the city loses substantial revenue when individuals don’t go through the process of filing a permit.”
Nashville spans roughly 500 square miles, making it impractical to monitor illegal road closures with cameras alone. Instead, Mukhopadhyay is applying AI and machine learning to analyze traffic speed data to identify potential closures. Whenever a road is blocked, traffic temporarily slows, allowing these algorithms to pinpoint locations that NDOT staff should inspect.
Mukhopadhyay joined the William & Mary Department of Computer Science this January and is in the process of building his own lab, the RAIL Lab: Robust and Adaptive Intelligence Lab, which will focus on developing AI techniques for large cyber-physical systems, such as transportation and emergency response.
“Not only will students contribute fundamental knowledge to the field, but they will also get to solve real-world problems,” Mukhopadhyay says. “We want to develop intelligence that is not for just one specific task, but rather adapts and improves as it collects data in a changing environment.”
In addition, Mukhopadhyay teaches a graduate-level course on mathematical models used in decision-making within AI and machine learning frameworks.
Through his research, teaching, and the development of the RAIL Lab, Mukhopadhyay is helping shape a generation of AI practitioners focused on practical applications. These efforts reflect William & Mary’s commitment to public service, applied scholarship, and research that addresses society’s most pressing challenges. His work moves AI beyond small-scale mathematical models and into decision-making contexts with tangible societal impact. As these systems evolve, their success will be measured by their ability to help people make smarter decisions when it matters most.
Eva Kalajian, School of Computing, Data Sciences & Physics
