Jinliang Wei
Staff Software Engineer at Google
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJinliang Wei is a Staff Software Engineer at Google with 13 years of experience building high-performance machine learning infrastructure, specializing in TPU, XLA, and large language model workflows. He holds a Ph.D. in Computer Science from Carnegie Mellon and has advanced compiler and distributed-training systems through both research and production roles. His open-source contributions to flagship projects like TensorFlow and OpenXLA include extending XLA's while-loop analysis and introducing asynchronous collective-permute opcodes, work that directly improves compiler optimizations and accelerator communication. Based in Mountain View, he blends deep systems thinking with practical engineering—evident from research internships and projects ranging from parallel SGD at Microsoft Research to runtime load balancing at HP Labs. Colleagues describe him as a meticulous backend engineer who uncovers subtle static-analysis and dataflow opportunities that yield measurable performance gains.
13 years of coding experience
12 years of employment as a software developer
Bachelor's Degree Computer Engineering minor in Mathematics, Bachelor's Degree Computer Engineering minor in Mathematics at Purdue UniversityDoctor of Philosophy (Ph.D.) Computer Science, Doctor of Philosophy (Ph.D.) Computer Science at Carnegie Mellon University
Chinese, English
Github Skills (20)
c-language10
dataflow10
operation10
machine-learning10
analyse10
compiler-design10
tensorflow10
hla10
xla10
cprogramming-language10
data-analysis10
python9
deep-learning9
optimization9
compiler8
Programming languages (4)
Github contributions (5)
openxla/xla
May 2020 - Dec 2022
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:
Back-end Developer
Back-end DeveloperContributions:26 commits in 2 years 6 months
Contributions summary:Jinliang contributed to the development of asynchronous collective-permute operations within the XLA compiler, introducing new HLO opcodes for asynchronous communication. Their work involved modifying the HLO verifier and instruction creation processes to support the new opcodes and ensure their correct usage. The user also implemented data flow analysis for the asynchronous collective-permute, fixing a bug and ensuring the proper handling of values. Furthermore, they added while-loop all-reduce code motion, optimizing the compiler's performance.
compilercommunity-drivenmachine-learningmodular
tensorflow/tensorflow
Mar 2020 - Dec 2022
An Open Source Machine Learning Framework for Everyone
Role in this project:Backend Developer
Backend DeveloperContributions:29 commits, 1 comment in 2 years 8 months
Contributions summary:Jinliang's commits focused on enhancing the XLA (XLA) compiler, particularly its analysis of while loops. They extended the pattern-match-based while loop analysis to handle cases with statically known loop bounds or increments, even when these cannot be evaluated within the while instruction. This involved improvements to the HLO evaluator to support parameter evaluation and expanding the analysis to accommodate multiple copy instructions within the loop patterns. These changes contribute to improving the compiler's optimization capabilities.
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.