Scholar
Aditi Krishnapriyan
Google Scholar ID: 7HoFN1wAAAAJ
Assistant Professor, UC Berkeley
Citations & Impact
Citations
1,990
H-index
17
i10-index
22
Publications
20
Co-authors
0
Contact
Publications
Resume
Academic Achievements
- 2025: Exploring emergent capabilities of pure Transformers on molecular data (arXiv:2510.02259)
- 2025: Accelerating molecular dynamics by repurposing generative models via statistical mechanics (ICML)
- 2025: Distilling large ML force fields into fast, physics-consistent models for molecular dynamics (ICLR)
- 2024: Principled scaling of neural interatomic potentials (NeurIPS)
- 2025: Improving stability and timestep length in neural potentials via differentiable Boltzmann estimators (TMLR)
- 2024: Neural operators with spectral methods and Parseval-based spectral loss for PDEs (ICLR)
- 2023–2024: PDE-constrained layers in neural networks, scaled via mixture-of-experts (ICLR)
- 2024: Work on building equivariance into neural networks (ICLR Spotlight)
- 2024: Publication in Journal of Chemical Information and Modeling (JCIM)
- Full publication list available on Google Scholar
Background
- Assistant Professor in Chemical Engineering and EECS at UC Berkeley
- Member of Berkeley AI Research (BAIR)
- Part of the AI+Science group in EECS and the theory group in Chemical Engineering
- Faculty scientist in the Applied Mathematics and Computational Research Division at LBNL
- Research focuses on developing machine learning methods driven by challenges in natural sciences, especially physics-inspired ML
- Key interests: physical inductive biases in learning, ML for scientific problems, enhancing physics-based solvers via differentiable frameworks, and handling distribution shifts in physical sciences
- Applications span atomistic and continuum domains, including fluid mechanics and molecular dynamics
- Interdisciplinary connections with numerical analysis, dynamical systems, quantum mechanics, computational geometry, optimization, and category theory