Jeffrey Daily
Fellow at AMD
Richland, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolJeffrey Daily is a Fellow at AMD with 11 years of experience bridging high-performance computing research and production-grade ML frameworks, focused on enabling AMD GPU support across projects like PyTorch and ONNX Runtime. He holds a Ph.D. in Computer Science from Washington State University and has published over 40 peer-reviewed papers while contributing foundational code to open-source projects such as Global Arrays, ComEx, Parasail, and FNCS. At AMD he advanced ROCm capabilities—optimizing kernels, hipifying CUDA code, and improving runtime stability and IPC—delivering measurable performance gains like 2x speedups on key operations. His open-source work also spans portability efforts (simde) and CI/tooling (pytorch/builder), reflecting a blend of low-level systems engineering, DevOps, and ML performance tuning. Based in Richland, WA, he combines deep academic rigor with pragmatic engineering to make cutting-edge GPU acceleration broadly accessible.
11 years of coding experience
19 years of employment as a software developer
Master of Science - MS Computer Science, Master of Science - MS Computer Science at Washington State University Tri-CitiesDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Washington State University
Github Skills (40)
pytorch10
github-ci10
docker10
c-language10
runtimes10
scripting10
memory-management10
machine-learning10
bash10
run-time10
onnx10
vectorization10
hardware-acceleration10
dockers10
cicd10
Programming languages (13)
Github contributions (5)
pytorch/pytorch
Mar 2019 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end & Performance Engineer
Back-end & Performance EngineerContributions:628 reviews, 272 commits, 342 PRs in 3 years 10 months
Contributions summary:Jeffrey primarily focused on improving the performance and stability of the PyTorch framework, particularly for ROCm (AMD) platforms. Their contributions involved optimizing CUDA kernels, specifically related to index_put operations, achieving a 2x performance boost. They also worked on enabling and improving the functionality of the hipblaslt library and addressing related issues to ensure feature parity with CUDA. In addition, the user was responsible for adding a check for new type enums.
pythongpu-accelerationdeep-learninggpunumpy
pytorch/builder
Sep 2020 - Aug 2022
Continuous builder and binary build scripts for pytorch
Role in this project:DevOps Engineer
DevOps EngineerContributions:16 reviews, 23 commits, 28 PRs in 1 year 11 months
Contributions summary:Jeffrey primarily contributed to the continuous integration and continuous delivery (CI/CD) aspects of the repository. Their work included adding and updating build scripts and Docker configurations, specifically targeting ROCm builds for the PyTorch project. They introduced support for various ROCm versions, including 3.7, 3.8, 3.9, 3.10, 4.0, 4.0.1, 4.1, and 4.2, and 5.2 by modifying build and deployment scripts. Moreover, the user implemented changes to the manywheel build process, including adjustments to dependencies and auxiliary file copying.
pytorchbuild-scriptscontinuousbuilder
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.