Anlun Xu
Software Engineer at Google
Cupertino, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAnlun Xu is a software engineer in Cupertino with 10 years of experience focused on ML compilers and high-performance GPU back ends, currently building production compiler/runtime features at Google. He has deep hands-on experience improving XLA and JAX performance—work that includes GPU/CPU compilation caching, command buffer compatibility, GPU AOT support, and robust custom-call error handling across prominent open-source projects like jax and TensorFlow. A Carnegie Mellon alumnus with top grades (BS 3.9, MS 4.0), he combines rigorous academic training with pragmatic systems engineering to ship low-level optimizations that meaningfully reduce ML training and inference costs. Notably, his contributions bridge compiler internals and runtime plumbing, making CUDA-related functionality and multi-stream GPU graph execution more reliable in production.
10 years of coding experience
Bachelor's degree, Computer Science, 3.9, Bachelor's degree, Computer Science, 3.9 at Carnegie Mellon University
Github Skills (22)
c-language10
gpu-programming10
machine-learning10
mlr10
triton10
tensorflow10
gpu10
performance-optimization10
compile10
cuda10
xla10
compiler10
jax10
cprogramming-language10
jit10
Programming languages (7)
Github contributions (5)
openxla/xla
Jul 2022 - Jan 2023
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:15 reviews, 71 commits, 16 comments in 6 months
Contributions summary:Anlun's contributions primarily focused on improving error handling within the XLA compiler, particularly within the runtime environment. They introduced a new mechanism for returning errors in custom call handlers for convolutions, launch functions, and GEMM operations. Further improvements involved the implementation of error messages in handlers for a wide array of custom calls (including AllReduce, Memset, Memcpy and others), enhancing the robustness of the XLA framework. Finally, the user made contributions towards GPU AOT compilation support.
compilercommunity-drivenmachine-learningmodular
tensorflow/tensorflow
Jul 2022 - Jan 2023
An Open Source Machine Learning Framework for Everyone
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:69 commits, 1 comment in 6 months
Contributions summary:Anlun's commits primarily focused on enhancing the GPU-based XLA compiler, specifically for machine learning tasks within the TensorFlow framework. The user added protobuf definitions for GPU compute capabilities, removed device ordinals from the GPU AOT pipeline, and incorporated features like GpuTargetConfigProto. They also improved the command buffer scheduling and implemented a more flexible and efficient approach to matrix-vector and matrix-matrix multiplication. Furthermore, they contributed to supporting and optimizing various HLO operations such as convolutions and also contributed to the implementation of runtime for GPU graph multi-streaming.
pythondata-sciencedeep-learningmlmachine-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.