Eugene Zhulenev
Software Engineer at Google
San Francisco, California, United States
Summary
🤩
RockstarEugene Zhulenev is a seasoned software engineer with 14 years of experience specializing in machine learning compilers, runtimes, and performance optimization. Based in San Francisco and currently working on ML compilers and runtimes at Google, he contributes deep systems-level expertise to projects like JAX, XLA, TensorFlow, IREE, and Eigen. His work focuses on backend performance—enabling persistent compilation caches, optimizing CPU runtimes, integrating CUDA support, and improving memory and kernel behavior for JIT runtimes. Eugene routinely tackles subtle low-level issues (ODR violations, thread-pool heuristics, and efficient block copies) that materially speed up ML workloads on CPUs and accelerators. He combines rigorous benchmarking with pragmatic engineering, shipping kernel and runtime changes that bridge research frameworks and production systems. An active open-source contributor, he brings both compiler architecture insight and hands-on performance engineering to large, widely used ML projects.
14 years of coding experience
Stackoverflow
9,734reputation
467kreached
182answers
14questions
sbt
top-5%
scala
top-1%
rdd
top-5%
intellij-idea
top-5%
apache-spark
top-1%
serialization
top-5%
Github Skills (54)
apache-spark10
c-language10
compilation10
python10
operation10
algebra10
llvm10
tensorrt10
machine-learning10
c1110
html-template10
python-templates10
mlr10
scala10
c1710
Programming languages (9)
Github contributions (5)
tensorflow/runtime
Apr 2020 - Dec 2022
A performant and modular runtime for TensorFlow
Role in this project:
Back-end Developer
Back-end DeveloperContributions:468 commits in 2 years 8 months
Contributions summary:Eugene primarily contributed to the core runtime and compilation aspects of the TensorFlow project, specifically focusing on the JitRt (Just-In-Time Runtime) and its associated kernel implementations. The commits reveal work on enhancing the handling of memory management, particularly for tensor operations, and integrating custom call functionality for improved performance. Furthermore, the user implemented several new native operations and also modified and improved existing kernels that are core to the runtime.
runtimeperformantmodulartensorflow
openxla/xla
Jan 2019 - Jan 2023
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:Back-end Developer
Back-end DeveloperContributions:462 reviews, 378 commits, 3 PRs in 4 years 1 month
Contributions summary:Eugene contributed to the XLA compiler project, with the primary focus on improving the runtime and the core components. Their work includes enabling support for custom contraction kernels in XLA's single-threaded matrix multiplication, implementing batch normalization through the cuDNN BatchNormEx API, and fixing ODR violations in Eigen contraction kernels. Furthermore, they added support for executing XLA:GPU on top of JitRt and improved the XLA runtime library.
compilercommunity-drivenmachine-learningmodular
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.