JZ-LIANG
Data Scientist
Beijing, China
Summary
🤩
RockstarData Scientist with 8 years of experience based in Beijing, specializing in distributed deep learning infrastructure and optimizer development. As a backend and ML engineer, contributed core optimizer implementations (LARS, LAMB) and distributed strategy integrations to the widely used PaddlePaddle framework, plus testing and 3D training support for large models. Also authored and improved technical documentation for distributed features like sharding, recompute-offload, and hybrid parallelism, bridging engineering and user-facing guidance. Comfortable navigating framework internals to deliver both production-ready code and clear developer-facing docs — a practical engineer who pairs low-level systems work with an eye for explainability.
8 years of coding experience
Github Skills (8)
paddlepaddle10
machine-learning10
distributed-training10
deep-learning10
python10
documentation10
unit-testing9
tensorflow6
Programming languages (5)
Github contributions (5)
PaddlePaddle/Paddle
Aug 2020 - Jan 2023
PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:688 reviews, 74 commits, 456 PRs in 2 years 5 months
Contributions summary:JZ-LIANG implemented and integrated the Lars and Lamb meta-optimizers within the PaddlePaddle framework, focusing on optimizing distributed deep learning models. They added configurations for these optimizers in the distributed strategy, enabling their use within distributed training setups. Furthermore, the user contributed to the testing suite, ensuring the correct application and behavior of the new optimization methods, and included the code related to the 3d training of bigtables. This involved modifying existing files and adding new files, demonstrating a strong understanding of the framework's internals and optimization techniques.
pytorchpythonparalleldeep-learningpaddlepaddle
PaddlePaddle/docs
Sep 2020 - Apr 2021
Documentations for PaddlePaddle
Role in this project:Technical Writer
Technical WriterContributions:1 review, 7 commits, 21 PRs in 7 months
Contributions summary:JZ-LIANG primarily contributed to the documentation of PaddlePaddle, specifically focusing on the distributed training features. They added and updated documentation for features such as LARS, LAMB optimizers, sharding, recompute-offload, and hybrid parallelism within the distributed strategy. Their work included examples, code snippets, and explanations to enhance the clarity and usability of the PaddlePaddle documentation.
deep-learningpaddlepaddledocumentationspaddlepaddle-tutorials
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.