MIT researchers and alumni selected as 2025 AI2050 Fellows
Two current MIT affiliates - postdoc Zongyi Li and Associate Professor Tess Smidt '12 - have been named AI2050 Early Career Fellows. They join seven MIT alumni recognized for work that pushes AI toward solving hard, practical problems in science and society.
The AI2050 program, conceived by Eric Schmidt and James Manyika and run by Schmidt Sciences, asks a simple prompt with real stakes: It's 2050 and AI is widely beneficial - what did we do to get there? The 2025 cohort reflects that focus, emphasizing research with measurable impact.
Why this cohort matters
Schmidt Sciences supports high-leverage research across AI and advanced computing, astrophysics, biosciences, climate, and space. The AI2050 Fellows program backs researchers building methods, tools, and theory with clear downstream use in science and engineering.
For working scientists, this means more rigorous methods, better scaling behavior, and tighter links between algorithms and physical constraints. Expect progress on simulation speedups, generalization across domains, and trustworthy deployment. Learn more about the initiative at Schmidt Sciences.
MIT fellows advancing core methods
Zongyi Li is a postdoc in CSAIL working with Associate Professor Kaiming He. Li focuses on neural operator methods that accelerate scientific computing - a path to faster PDE solvers, reduced training cost, and broader applicability in physics-driven models. He earned his PhD at Caltech with advisors Anima Anandkumar and Andrew Stuart, holds undergraduate degrees in computer science and mathematics from Washington University in St. Louis, and has been supported by the Kortschak Scholarship, PIMCO Fellowship, Amazon AI4Science Fellowship, Nvidia Fellowship, and MIT-Novo Nordisk AI Fellowship. Li will join the NYU Courant Institute as an assistant professor of mathematics and data science in fall 2026. More on his lab's ecosystem at MIT CSAIL.
Tess Smidt '12, associate professor of EECS, leads the Atomic Architects group at the Research Laboratory of Electronics (RLE). Her work sits at the intersection of physics, geometry, and machine learning, with a specific focus on symmetries in 3D systems - rotation, translation, and reflection - to design algorithms for new materials and molecule discovery. She received her BS in physics from MIT and PhD from UC Berkeley, served as the 2018 Alvarez Postdoctoral Fellow in Computing Sciences at Lawrence Berkeley National Laboratory, and worked as a software engineering intern on the Google Accelerated Sciences team, where she developed Euclidean symmetry equivariant neural networks for 3D geometry and tensor data. Honors include an AFOSR Young Investigator award, the EECS Outstanding Educator Award, and a Transformative Research Fund award.
2025 AI2050 Fellows with MIT ties
Early Career Fellows
- Zongyi Li (MIT CSAIL postdoc)
- Tess Smidt '12 (MIT EECS associate professor)
- Brian Hie SM '19, PhD '21
- Natasha Mary Jaques PhD '20
- Martin Anton Schrimpf PhD '22
- Lindsey Raymond SM '19, PhD '24 (joining MIT EECS, Economics, and MIT Schwarzman College of Computing in 2026)
- Ellen Dee Zhong PhD '22
Senior Fellows
- Surya Ganguli '98, MNG '98
- Luke Zettlemoyer SM '03, PhD '09
What to watch for in the research
Li's neural operator work points to faster, more accurate scientific simulations - especially for PDE-driven domains where training data is expensive. That can shorten experiment cycles, improve uncertainty estimates, and make large-scale parameter sweeps feasible.
Smidt's symmetry-aware models continue to tighten the link between physical law and learning. Expect methods that respect invariances out of the box, reduce sample complexity, and transfer across molecular and materials tasks without breaking physical constraints.
Across the cohort, look for open-source releases, benchmark improvements grounded in physics and data efficiency, and cross-lab collaborations tackling compute bottlenecks and real-world deployment. The north star remains clear: solve the hardest problems with AI while keeping results useful, verifiable, and safe.
Upskilling for scientists
If you're aligning your research with these directions and need structured learning paths, explore curated options by role at Complete AI Training.
Your membership also unlocks:
