Anandkumar, the Bren Professor of computing and mathematical sciences at Caltech, has pioneered an innovative approach to artificial intelligence that bridges theoretical frameworks with practical applications. Her work focuses on creating algorithms rooted in fundamental principles derived from mathematics and physics, a method she developed during a time when deep learning was not yet prominent. Anandkumar, who has also held significant roles at Amazon Web Services and Nvidia, emphasizes the need for algorithms to be “more principled, hardware efficient, and robust.”
Her collaboration has led to the creation of “neural operators,” a universal AI framework that can learn to simulate physical processes across varying scales—ranging from molecular interactions to broader climate patterns. Unlike traditional AI models like ChatGPT, these models integrate the laws of physics, allowing them to assess the plausibility of their predictions. Anandkumar highlights that this capability enables the models to “learn shortcuts” based on their training data, significantly improving simulation efficiency and accuracy while reducing the need for extensive computational resources.
In 2022, Anandkumar’s interdisciplinary team from Nvidia, Caltech, and other institutions developed FourCastNet, a revolutionary AI-driven open-source weather model utilizing neural operators. This model demonstrated a processing speed tens of thousands of times faster than conventional numerical weather prediction systems, while often enhancing accuracy. Within less than two seconds, FourCastNet can generate a week-long forecast across various variables, such as wind speed and precipitation—an achievement that previously required supercomputers and several hours of processing time. The model is accessible online via the European Centre for Medium-Range Weather Forecasts and has garnered attention for improving extreme weather forecasts, as highlighted by Anandkumar during the accurate prediction of Hurricane Beryl’s trajectory in June 2024.
Further demonstrating the versatility of her neural operators, Anandkumar’s team partnered with the U.K. Atomic Energy Agency in 2024 to simulate plasma behavior within nuclear fusion reactors at speeds over a million times faster than previous methods. This advancement is critical for predicting and preventing plasma disruptions, volatile occurrences that can jeopardize reactor integrity, thereby enabling technicians to implement corrective measures proactively.
Anandkumar’s innovations extend beyond predictive modeling; they also play a significant role in design applications. In 2023, she collaborated with Caltech researchers to develop a catheter aimed at reducing catheter-associated urinary tract infections, which affect over a million Americans annually. By simulating fluid flow, the AI identified optimal locations for tiny grooves within the catheter that hinder bacterial movement, achieving a hundred-fold reduction in contamination potential. This adaptable AI framework can evaluate and refine various designs across disciplines, including medical devices and drug development.
Looking ahead, Anandkumar envisions a future where AI and scientific inquiry mutually reinforce each other, creating systems that integrate scientific understanding with AI capabilities. “Many labs, including us, are building towards this,” she stated, emphasizing the ongoing discoveries in the field. Her work highlights the potential for AI to not only extend our understanding of the physical world but also to innovate in ways that were previously unimaginable.
This profile is part of TIME’s TIME100 Impact Awards initiative, which recognizes global leaders who are effecting change in their communities. The upcoming TIME100 Impact Awards ceremony is scheduled for February 10 in Dubai.
See also
IDF Launches AI-Driven Robots and Advanced Artillery Against Hezbollah’s Infrastructure
AI Awareness Reveals Dual Impact on Hotel Employees: Performance Boost or Job Stress?
IGM Library and Turnitin Launch Training on Academic Integrity Amid AI Challenges
Qualcomm Reveals 3D DRAM NPU Project for AI Phones, Promising 40 TOPS by 2027
Study Finds Elon Musk’s Grok Most Dangerous AI Model for Reinforcing Delusions
