Anima Anandkumar is Caltech’s Bren Professor of Computing and Mathematical Sciences and a cofounder of Accelerated Understanding. She develops AI methods for scientific simulation, including neural operators that learn transformations between physical fields.[1][2][4]
Start with Fourier Neural Operator for Parametric Partial Differential Equations →
Education
Links
Research site CVCofounder · Accelerated Understanding
Bren Professor · California Institute of Technology
Career
Her career spans postdoctoral research at MIT, faculty appointments at UC Irvine, Amazon AI and NVIDIA research leadership before her current academic and company work.[2]
Contributions
The Fourier Neural Operator paper learns a mapping between functions, such as an input coefficient field and the solution of a partial differential equation. Learned integral kernels are parameterized in Fourier space. This permits shared parameters across discretizations instead of training an unrelated network for each grid. The paper evaluates Burgers, Darcy and Navier–Stokes problems; its speed and accuracy results concern those experiments, rather than every physical simulation. Anandkumar is one of the paper’s coauthors.[4]
Ideas
Her Latent Space interview connects operator learning with weather, fluid dynamics and other continuous systems. A field has structure across space and resolution that a flat sequence of tokens may represent inefficiently. Spherical harmonics provide a natural basis for global weather dynamics. The discussion also covers neural-network verification and the longer-term aim of models that support both scientific simulation and design.[5]
Tools, writing & talks
Writing, research & talks
Start here
The original paper on learning solution operators with Fourier-space kernels.
Start with the paper’s operator-learning formulation and experimental figures. Linear algebra, Fourier transforms and basic differential equations help; the interview provides a less mathematical companion.
Coauthor[4]- 🔬“We have foundation models for language, not for physics” — Anima Anandkumar, Bren Professor of Computinglatent.space
A Latent Space interview about neural operators, physical structure, weather and scientific discovery.
Interview guest[5]Original source