“`html
For Priya Donti, childhood excursions to India represented more than just a chance to connect with extended relatives. The biannual travels ignited within her a drive that continues to influence her research and pedagogy.
In stark contrast to her family residence in Massachusetts, Donti — currently the Silverman Family Career Development Professor in the Department of Electrical Engineering and Computer Science (EECS), a shared role between the MIT Schwarzman College of Computing and EECS, and a lead investigator at the MIT Laboratory for Information and Decision Systems (LIDS) — was impacted by the differences in living conditions.
“It became abundantly clear to me how pervasive inequality is globally,” Donti remarks. “From an early age, I was determined to tackle that challenge.”
This drive was further fueled by a high school biology instructor, who concentrated his lessons on climate and sustainability.
“We discovered that climate change, this monumental and urgent issue, would intensify inequality,” Donti states. “That stayed with me and ignited a passion within me.”
Thus, when Donti enrolled at Harvey Mudd College, she intended to channel her enthusiasm into chemistry or materials science to develop advanced solar panels.
However, those ambitions took an unexpected turn. Donti “fell in love” with computer science and soon encountered research from the United Kingdom advocating that artificial intelligence and machine learning would be vital in integrating renewable energy into power grids.
“It was the first moment I had seen those two passions united,” she recalls. “I became captivated and have pursued that subject ever since.”
While pursuing her PhD at Carnegie Mellon University, Donti tailored her program to incorporate computer science and public policy. In her research, she investigated the necessity for fundamental algorithms and tools capable of managing power grids that heavily depend on renewables at scale.
“I aimed to play a role in developing those algorithms and toolkits by formulating new machine learning methods grounded in computer science,” she explains. “But I also wanted to ensure that my approach was rooted in the actual energy systems sector and collaborating with individuals from that sector” to offer what was genuinely required.
During her PhD studies, she co-founded a nonprofit organization named Climate Change AI. Her mission, she states, was to assist the community of those involved in climate and sustainability — “whether they are computer scientists, researchers, practitioners, or policymakers” — to unite and access resources, connections, and education “to aid them on that journey.”
“Within the climate sector,” she notes, “you require specialists in specific climate change-related fields, as well as experts in diverse technical and social science tools, problem stakeholders, impacted users, and policymakers who understand the regulations — all these elements — to have an on-the-ground scalable impact.”
Upon arriving at MIT in September 2023, it was no surprise that she was attracted by its initiatives focused on applying computer science to tackle society’s most pressing issues, particularly the ongoing threat to planetary health.
“We’re genuinely contemplating where technology has a long-term impact and how technology, society, and policy must collaborate,” Donti explains. “Technology isn’t merely a one-off project that can be monetized within a year.”
Her research employs deep learning models to integrate the physics and strict constraints of electric power systems utilizing renewables for improved forecasting, optimization, and control.
“Machine learning is already extensively applied to tasks like solar power forecasting, which is essential for managing and balancing power grids,” she states. “My emphasis lies in how to enhance the algorithms for genuinely balancing power grids in the face of varying renewable energy sources over time.”
Among Donti’s innovations is a promising method enabling power grid operators to optimize costs by considering the actual physical conditions of the grid, rather than depending on estimates. Although the solution hasn’t been implemented yet, it appears to operate ten times quicker and at a significantly lower cost than previous technologies, and has garnered interest from grid operators.
Another technology she is developing aims to generate data that can be utilized in training machine learning systems for power system optimization. Generally, much of the data related to these systems is confidential, either due to proprietary reasons or security concerns. Donti and her research team are focused on creating synthetic data and benchmarks that, as Donti states, “can help illuminate some of the underlying issues” in making power systems more efficient.
“The question is,” Donti asks, “can we elevate our datasets to a level where they are just challenging enough to stimulate progress?”
In recognition of her endeavors, Donti has received the U.S. Department of Energy Computational Science Graduate Fellowship and the NSF Graduate Research Fellowship. She was acknowledged in MIT Technology Review’s 2021 listing of “35 Innovators Under 35” and Vox’s 2023 “Future Perfect 50.”
Next spring, Donti will co-instruct a course entitled AI for Climate Action alongside Sara Beery, EECS assistant professor focused on AI for biodiversity and ecosystems, and Abigail Bodner, assistant professor in EECS and Earth, Atmospheric, and Planetary Sciences, who concentrates on AI for climate and Earth science.
“We’re all incredibly enthusiastic about it,” Donti shares.
Upon joining MIT, Donti expressed, “I realized there would be a network of individuals who genuinely cared, not just about success indicators like publications and citation metrics, but about the societal impact of our work.”
“`