Zhengxiao Du
Judge S Associate At District Court Of New South Wales
Australia
Summary
🤩
Rockstar🎓
Top SchoolZhengxiao Du is a legal professional and technologist with nine years of experience who currently serves as a Judge’s Associate at the District Court of New South Wales. Trained in law with degrees from the University of Sydney and international legal study in China, he couples courtroom procedure and legal research with practical experience as a court officer, paralegal and legal clerk. Unusually for a courtroom role, he also contributes to leading open-source AI projects—implementing and refining generation and inference components in high-profile repositories such as THUDM’s GLM and GLM-130B—bringing machine learning engineering skills to bear on complex model and training pipelines. He has hands-on experience with model parallelism, DeepSpeed training setups, generation strategies like beam search, and tokenizer-agnostic data handling, signaling a rare cross-disciplinary fluency. Based in Australia, Zhengxiao blends rigorous legal judgment with applied ML engineering, making him adept at translating technical complexity into reliable, auditable processes for high-stakes environments.
9 years of coding experience
Master of Professional Accounting, Accounting and Finance, Master of Professional Accounting, Accounting and Finance at Top Education Group LtdThe University of SydneyBachelor of Laws - LLB, Law, Bachelor of Laws - LLB, Law at University of Electronic Science and Technology of China
Github Skills (16)
transformer-models10
model-building10
beam-search10
machine-learning10
pytorch10
code-generation10
nlp10
deepspeed10
python10
natural-language-processing10
modeling10
pre-trained-model9
parallelization9
tokenizer8
data-loading8
Programming languages (2)
Github contributions (5)
THUDM/GLM
Oct 2020 - Jan 2023
GLM (General Language Model)
Role in this project:
Back-end Developer
Back-end DeveloperContributions:825 commits, 6 PRs, 82 pushes in 2 years 3 months
Contributions summary:Zhengxiao primarily contributed to the development of the GLM (General Language Model) repository by modifying the data loading, tokenization, and generation components. They introduced the ability to use the model without a tokenizer, improved the training procedure, and enhanced the handling of dataset paths and filepaths. Moreover, the user made changes to support the dataset and model parallelism setup for training, specifically involving the use of DeepSpeed for model training.
glmlanguage-modelmultimodaldialogue-systems
THUDM/SwissArmyTransformer
Oct 2021 - Nov 2022
SwissArmyTransformer is a flexible and powerful library to develop your own Transformer variants.
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:61 commits, 1 branch in 1 year 1 month
Contributions summary:Zhengxiao primarily contributed to the implementation of the GLM model within the SwissArmyTransformer library. Their work involved adding the GLM model and importing the necessary inference code for the model. They also made several fixes to the attention mask in the GLM model, correcting some potential errors in the inference. Furthermore, the user made some changes to the training/model_io.py.
pytorchpretrained-modelstransformerflexible
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.