Roy Hvaara
Software Engineer at Arundo Analytics, Inc.
San Francisco, California, United States
Summary
👤
Senior🎓
Top SchoolRoy Hvaara is a software engineer based in San Francisco with 11 years of experience building reliable, production-grade systems across SRE and machine learning projects. He spent several years on Google’s Site Reliability Engineering team and previously led engineering efforts at Arundo Analytics and other firms, combining infrastructure scaling, ops, and backend development skills. Roy is an active open-source contributor in major ML projects—most notably Transformers and PyTorch—where he improved TensorFlow/JAX integrations and advanced Metal Performance Shaders support for GPU acceleration. He brings a practical fusion of ML engineering and systems reliability, able to move models from research into robust, performant deployments. Unusually, his background spans informatics and early medical studies, reflecting a broad analytical foundation that informs his pragmatic, interdisciplinary approach to problem solving.
11 years of coding experience
9 years of employment as a software developer
Bachelor’s Degree Informatics: Programming and Networks, Bachelor’s Degree Informatics: Programming and Networks at University of OsloMedicine, Medicine at University of PécsPhysics Biology Math and Chemistry, Physics Biology Math and Chemistry at Re videregåendeCoursera
English, Norwegian
Stackoverflow
139reputation
4kreached
5answers
0questions
Github Skills (27)
pytorch10
python10
metal-performance-shaders10
machine-learning10
deep-learning10
tensorflow10
gpu10
neural-network10
tensor10
jax10
mps10
numpy9
flax9
autograd9
test-automation9
Programming languages (9)
Github contributions (5)
pytorch/pytorch
Apr 2024 - Feb 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:23 reviews, 19 PRs, 228 comments in 10 months
Contributions summary:Roy primarily contributed to improving the PyTorch framework's support for Metal Performance Shaders (MPS), focusing on optimizations, bug fixes, and new feature implementations related to tensor operations and neural network modules. They addressed issues with boolean data types, batch normalization with sliced inputs, and the `fftfreq` function. Furthermore, the user added and updated regression tests to ensure the correctness of MPS implementations, especially regarding `nn.Conv3d` and `F.linear`. Their work also involved refining error messages and adding autocast support.
pythongpu-accelerationdeep-learninggpunumpy
huggingface/transformers
Jan 2023 - Jan 2023
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:1 review, 1 commit, 9 PRs in 1 day
Contributions summary:Roy primarily contributed to the Hugging Face Transformers library, making various improvements related to the TensorFlow and JAX integrations. Their work included refactoring code to remove deprecated NumPy type aliases, adding support for TensorFlow's `is_symbolic_tensor` predicate, and replacing instances of `jnp.array` with `jnp.ndarray` in JAX code. They also updated dependencies to use the latest Keras initializers.
pythonbertspeech-recognitionstate-of-the-artflax
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.