For Priya Donti, her childhood trips to India were more than just an opportunity to visit extended family. Her two-year travels sparked a drive that continues to shape her research and teaching.
Comparing her childhood home in Massachusetts, Donti—now the Silverman Family Career Development Professor at MIT's Department of Electrical Engineering and Computer Science (EECS) and a principal investigator at MIT's Information and Decision Systems Laboratory—was struck by the disparities in the way people lived.
“It was very clear to me how much inequality is a widespread problem around the world,” Donti says. “I knew from a very early age that I definitely wanted to address this issue.”
This motivation was further fueled by a high school biology teacher who focused his classes on climate and sustainability.
“We have learned that climate change – this huge and important problem – will worsen inequality,” Donti says. “That really stuck with me and lit a fire in my belly.”
So when Donti enrolled at Harvey Mudd College, she thought she would channel her energy into studying chemistry or materials science to create the next generation of solar panels.
However, these plans were abandoned. Donti “fell in love” with computer science and then discovered the work of researchers in the UK who argued that artificial intelligence and machine learning would be necessary to help integrate renewable energy sources into energy grids.
“It was the first time I saw a combination of these two interests,” she says. “I got hooked and have been working on it ever since.”
While studying for her PhD at Carnegie Mellon University, Donti was able to design her degree to include computer science and public policy. Her research examined the need for fundamental algorithms and tools that could manage large-scale energy networks that rely heavily on renewable energy sources.
“I wanted to be involved in developing these algorithms and toolkits, creating new computer science-based machine learning techniques,” she says. “But I wanted to make sure that the way I was doing this work was based both on the actual field of energy systems and on working with people in the field” to provide what was actually needed.
While Donti was working on her PhD, she co-founded a nonprofit organization called Climate Change AI. Her goal, she says, was to help the community of people involved in climate and sustainability – “whether they are computer scientists, scientists, practitioners or policymakers” – come together and access resources, connections and education “to help them on this journey.”
“In the climate space,” he says, “you need experts from specific climate change sectors, experts in different technical and social science toolkits, issue owners, impacted users, regulatory-savvy policymakers – all of them – to have a scalable impact on the ground.”
When Donti arrived at MIT in September 2023, it was no surprise that she was drawn to initiatives that direct the application of computer science to society's biggest problems, especially the current threat to the health of the planet.
“We're really thinking about where technology has a much longer-term impact and how technology, society and policy need to work together,” Donti says. “Technology is not a one-time thing and you can make money on it in one year.”
His work uses deep learning models to take into account the physics and hard constraints of renewable power systems for better forecasting, optimization and control.
“Machine learning is already very widely used for solar energy forecasting, which is a prerequisite for managing and balancing energy networks,” he says. “My focus is on how do we improve algorithms for actually balancing power grids in the face of a range of time-varying renewable energy sources?”
Donti's breakthroughs include a promising solution for power grid operators to optimize costs by taking into account the actual physical realities of the grid, rather than relying on approximations. Although the solution has not yet been implemented, it appears to be 10 times faster and much cheaper than previous technologies, which has attracted the attention of network operators.
Another technology it is developing aims to provide data that can be used to train machine learning systems to optimize the power system. Generally, much of the data associated with systems is private, either because it is proprietary or for security reasons. Donti and her research group are working to create synthetic data and patterns that Donti says “can help expose some of the fundamental problems” in making power systems more efficient.
“The question is,” Donti says, “can we get our datasets to the point where they are hard enough to drive progress?”
For her efforts, Donti was awarded a U.S. Department of Energy Computational Science Graduate Fellowship and an NSF Graduate Research Fellowship. She was considered part of it MIT Technology Reviewnamed to the 2021 “35 Innovators Under 35” list and to the Vox “Future Perfect 50” list in 2023.
Next spring, Donti will co-teach a class called AI for Climate Action with Sara Beery, an EECS assistant professor who focuses on artificial intelligence for biodiversity and ecosystems, and Abigail Bodner, an assistant professor of earth, atmospheric and planetary sciences who holds a joint appointment at the MIT Schwarzman College of Computing with EECS.
“We're all very excited about it,” Donti says.
Coming to MIT, Donti says, “I knew there would be an ecosystem of people who really cared not only about success metrics like publications and citation counts, but also about the impact of our work on society.”
