Oguz Ulgen
Principal Software Engineer at Meta
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolOguz Ulgen is a Principal Software Engineer with 11 years of experience specializing in compilers, language design, and type theory, currently working at Meta on the PyTorch compiler and Helion. He was a primary engineer on HHVM’s JIT compiler and contributes to high-profile open-source projects like PyTorch and Flow, bringing deep systems-level expertise in C++, Rust, and OCaml. His work spans production compiler backends, performance-critical Triton kernel integration, and infrastructure to speed up model compilation and caching. Based in Menlo Park, he combines research-minded rigor from a Carnegie Mellon CS background with practical impact at scale, and has a track record of improving server reliability and developer tooling in large, distributed codebases. An unusual thread through his career is applying formal language and type theory knowledge directly to ML compiler engineering, bridging theory and production performance.
11 years of coding experience
2 years of employment as a software developer
Bachelor of Science (B.S.), Computer Science, Bachelor of Science (B.S.), Computer Science at Carnegie Mellon UniversityHigh School Diploma, High School, High School Diploma, High School at Uskudar American Academy
English, Turkish
Stackoverflow
35reputation
6kreached
0answers
2questions
Github Skills (29)
caching10
pytorch10
sqlite10
python10
compiler-development10
machine-learning10
triton10
gpu10
compile10
model-optimization10
cuda10
tensor10
jit10
deep-learning9
logging9
Programming languages (8)
Github contributions (5)
pytorch/pytorch
Jun 2020 - Jun 2020
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
ML Engineer
ML EngineerContributions:716 reviews, 1 commit, 231 PRs in 1 day
Contributions summary:Oguz primarily contributed to the development and integration of Triton kernels within the PyTorch framework, specifically focusing on accelerating machine learning operations. Their work involved implementing, testing, and optimizing custom Triton kernels, addressing issues related to numerical stability and multi-dimensional indexing. The contributions demonstrate a deep understanding of performance considerations when integrating custom kernels, optimizing the performance of the kernels, and implementing the underlying data structures and mechanisms. The user also addressed issues related to the integration of these kernels within the framework.
pythongpu-accelerationdeep-learninggpunumpy
facebook/flow
Aug 2016 - May 2018
Adds static typing to JavaScript to improve developer productivity and code quality.
Role in this project:Back-end Developer
Back-end DeveloperContributions:39 commits in 1 year 8 months
Contributions summary:Oguz primarily contributed to the `facebook/flow` repository by implementing and modifying core backend functionalities related to server monitoring and logging. They added features to upload crash and error logs, extended the logging capabilities to include more log files, and integrated the use of sqlite for storing the deptable for faster initialization. Further, they worked on the server initialization processes by waiting for repository changes, and classified SQL assertion failures, along with implementing SQL optimizations.
static-typingproductivityjavascriptdeveloper-productivitydataflow-programming
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.