Zheng Xu
Texas, United States
Summary
🤩
RockstarZheng Xu is a Staff Software Engineer and Tech Lead Manager based in Mountain View with 12 years of experience building large-scale deep learning systems for video understanding, now leading Videos in Gemini at Google DeepMind. He combines research rigor from his Ph.D. in large-scale optimization with hands-on engineering, shipping production-ready video models and scalable training frameworks throughout Google Research. Zheng is an active open-source contributor to high-profile projects like YouTube-8M and MXNet, improving dataset readers, device workload balancing, and GPU compatibility for real-world training. His background spans ML engineering, data pipeline design, and model benchmarking—including contributions to DeepChem for molecular property prediction—reflecting a rare blend of video-centric expertise and cross-domain ML tooling. Colleagues describe him as passionate and driven to “discover and soar,” bringing both technical depth and practical impact to large distributed ML systems.
12 years of coding experience
Github Skills (20)
context-manager10
drug-discovery10
benchmark10
python10
evaluation10
data-science10
mxnet10
machine-learning10
benchmarking10
context-api10
eval10
deep-learning10
trainings10
tensorflow10
modeling10
Programming languages (14)
Github contributions (5)
google/youtube-8m
May 2019 - Oct 2019
Starter code for working with the YouTube-8M dataset.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:92 commits, 31 PRs, 20 pushes in 5 months
Contributions summary:Zheng primarily worked on refactoring and improving the YouTube-8M dataset's reading functionality. Their contributions included removing warnings, and refactoring the reader to support label_weight output. Additionally, the user addressed issues related to the computational graph by removing an assert node. They also contributed to fixing issues related to running the code on RTX series GPUs.
deep-learningdatasetmachine-learningyoutubeyoutube-8m
deepchem/deepchem
Mar 2017 - Mar 2017
Democratizing Deep-Learning for Drug Discovery, Quantum Chemistry, Materials Science and Biology
Role in this project:ML Engineer & Data Scientist
ML Engineer & Data ScientistContributions:12 commits, 3 PRs, 30 comments in 23 days
Contributions summary:Zheng's commits primarily involve modifications to the benchmark functionality within the deepchem repository. They refactored the benchmark design and introduced new datasets, specifically related to molecular property prediction. The user also made changes related to the featurizer and dataset loading, indicating involvement in the data preparation and model evaluation pipeline within the context of deep learning for chemistry. A small change related to progressive multitask models indicates some interest in the modeling side.
deep-learningquantum-chemistrydrug-discoverybiologymaterials-science
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.