Zheng Xu
Staff Research Scientist at Meta
United States
Summary
🤩
Rockstar🎓
Top SchoolZheng Xu is a staff research scientist with 11 years of experience building and deploying privacy-preserving machine learning systems, currently leading large-model research at Meta after a multi-year research leadership track at Google. His work bridges federated learning, differential privacy, and synthetic data—delivering production-grade DP guarantees (including ε < 1 on-device models) and shipping federated Gboard language models with user-level protections. He’s an active open-source contributor to flagship projects like TensorFlow Federated and TensorFlow Privacy, implementing practical tree-aggregation and FedAvg integrations that enabled real-world federated training on non-iid data. Trained at University of Maryland and USTC, he combines deep theoretical background in optimization with hands-on systems engineering, and has a track record of turning cutting-edge privacy research into scalable production systems.
11 years of coding experience
11 years of employment as a software developer
Master Information Science, Master Information Science at University of Science and Technology of ChinaDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of Maryland
Github Skills (14)
algorithm10
mask-rcnn10
data-structures10
keras10
algorithms10
faster-rcnn10
machine-learning10
privacy10
tensorflow10
federated-learning10
python10
data-structure10
data-science9
git8
Programming languages (4)
Github contributions (5)
google-research/federated
Jul 2020 - Dec 2022
A collection of Google research projects related to Federated Learning and Federated Analytics.
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:20 reviews, 89 commits, 7 PRs in 2 years 5 months
Contributions summary:Zheng migrated research projects related to Federated Learning and Federated Analytics from the `tensorflow/federated` repository to the `google-research/federated` repository. The commits demonstrate work on the `adaptive_lr_decay` functionality, including modifications to the `adaptive_fed_avg.py`, `adaptive_fed_avg_test.py`, `callbacks.py`, `callbacks_test.py`, and `federated_trainer.py` files. The user also made changes to the dataset construction, by sampling from Dirichlet distribution and modifying dataset util and training files. These changes indicate contributions in the core federated learning framework.
analyticsfederated-analyticsmachine-learningkubernetesfederated-learning
google-parfait/tensorflow-federated
Jan 2020 - Nov 2022
An open-source framework for machine learning and other computations on decentralized data.
Role in this project:ML Engineer
ML EngineerContributions:141 commits, 9 PRs, 3 pushes in 2 years 10 months
Contributions summary:Zheng implemented and tested federated learning models for the EMNIST dataset within the TensorFlow Federated framework. Their primary contribution involved integrating a CNN model and creating a Keras model wrapper to be used within the federated learning setup. Further, the user refactored the training process, ensuring the compatibility of the code with the TFF framework, which involved modifying client and server updates to work with the new `tff.learning.Model` and the adoption of TFF optimizers for the FedAvg algorithm.
pytorchdeep-learningmachine-learningsecure-computationfederated-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.