Simon Batzner
Senior Research Scientist at Google DeepMind
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSimon Batzner is a Staff Research Scientist at Google DeepMind with nine years of experience building high-performance ML systems that bridge reinforcement learning, large language models, and physics-informed modeling. He earned a PhD in Applied Math from Harvard, where he invented equivariant interatomic potentials and scaled them to run inference on 5,120 GPUs—work that was a finalist for the 2023 Gordon Bell Prize. Simon contributes to notable open-source projects like e3nn and NequIP, improving numerical basis functions, equivariant tensor computations, and experiment tracking for large-scale training. Based in San Francisco, he combines deep theoretical insight with hands-on engineering for supercomputing and production ML workflows, and has a track record of turning symmetry-aware research into robust, scalable code.
9 years of coding experience
8 years of employment as a software developer
SM, Computational Engineering, SM, Computational Engineering at Massachusetts Institute of TechnologyPhD, Applied Math, PhD, Applied Math at Harvard UniversityBSc, BSc at University of Stuttgart
French, English, German
Stackoverflow
133reputation
28kreached
2answers
1question
Github Skills (24)
pytorch10
python10
machine-learning10
deeplearning-ai10
deep-learning10
neural-network10
atomics9
simulate9
simulation9
simulations9
harmonic9
trainings9
numerical-methods9
modeling9
atomic9
Programming languages (4)
Github contributions (5)
mir-group/nequip
Mar 2021 - Nov 2022
NequIP is a code for building E(3)-equivariant interatomic potentials
Role in this project:
ML Engineer
ML EngineerContributions:19 reviews, 98 commits, 27 PRs in 1 year 8 months
Contributions summary:Simon primarily contributed to the development of the NequIP interatomic potential code. They introduced and updated trainer functionality for integration with Weights & Biases (WandB) for experiment tracking and logging. Additional contributions include code modifications related to data handling, model training, and evaluation, along with fixes and improvements to various model components, including energy calculations and interaction blocks.
pytorchinteratomic-potentialsmolecular-dynamicsdrug-discoverydeep-learning
e3nn/e3nn
Feb 2021 - Mar 2021
A modular framework for neural networks with Euclidean symmetry
Role in this project:ML Engineer
ML EngineerContributions:9 commits, 3 PRs, 2 pushes in 1 month
Contributions summary:Simon primarily contributed to the core functionality of the `e3nn` library, which focuses on neural networks with Euclidean symmetry. Their work included implementing features like a "bessel" basis within the `soft_one_hot` function, fixing edge cases, and improving documentation. The user's commits also touched on spherical tensor calculations, adding checks and correcting issues to ensure correctness. Their contributions focused on numerical methods and basis function implementations within the context of neural network design.
modular-frameworksymmetrydeep-learningneural-networkseuclidean
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.