Roy Frostig
Research Scientist at google-deepmind
sfba, United States
Summary
🤩
RockstarRoy Frostig is a research scientist and seasoned ML engineer in the San Francisco Bay Area with 16 years of experience building high-performance machine learning systems and compiler tooling. He is a JAX co-author and prolific open-source contributor, with substantive back-end work on foundational projects such as Stanford CoreNLP, XLA, JAX, JAXOPT, and TensorFlow Probability. His contributions span creating Python bindings for a compiler, implementing convolution and all-to-all operations, and advancing differentiable optimizers and quadratic-program solvers—demonstrating deep expertise in numeric optimization, compiler interfaces, and scalable ML infrastructure. Roy blends research rigor with production-driven engineering, often refactoring core classifier and shape-representation code to improve performance and flexibility. Notably, his history includes both low-level system work (XLA/JAX) and higher-level probabilistic and optimization tooling, giving him a rare full-stack perspective on ML stacks.
16 years of coding experience
Github Skills (21)
python10
statistics10
machine-learning10
java10
numpy10
javas10
deep-learning10
tensorflow10
natural-language-processing10
xla10
compiler10
differentiable-programming10
nlp10
jax10
linear-algebra10
Programming languages (9)
Github contributions (5)
jax-ml/jax
May 2020 - Dec 2022
Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:1106 reviews, 198 commits, 320 PRs in 2 years 7 months
Contributions summary:Roy's commits primarily involve source code synchronization across different repositories and contain additions of new functionality. The user's contributions include introducing and modifying several examples related to machine learning, such as implementing a kernel least-squares regression example, and running and testing ResNet50. The user's work also encompasses various JAX-related tasks like setting up custom random number generators for code use.
pytorchpythonjitautomatic-differentiationgpu
tensorflow/probability
Nov 2018 - Aug 2022
Probabilistic reasoning and statistical analysis in TensorFlow
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:322 commits in 3 years 9 months
Contributions summary:Roy contributed primarily to the core functionality of the TensorFlow Probability library. Their commits focused on improvements and synchronization within the codebase. Specific changes included updating setup scripts, integrating new features in statistical computations, and the addition of examples showcasing the library's capabilities, as well as general code maintenance and upgrades to dependencies.
statisticspythonprobabilistic-reasoningdata-sciencedeep-learning
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.