Shuang Song
Research Software Engineer at Google
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolShuang Song is a research software engineer based in the San Francisco Bay Area with six years of industry experience and a PhD in Computer Science from UC San Diego. At Google since 2018, she applies rigorous research methods to build production-ready ML systems, with a focus on privacy-preserving machine learning. Her open-source contributions to tensorflow/privacy include implementing membership inference attack tooling for both Estimator and Keras workflows, helping quantify privacy risks in widely used ML pipelines. With a foundation in math and computer science from HKUST, she blends theoretical depth and practical engineering to evaluate model vulnerability and measurement. Colleagues know her for translating complex privacy research into reusable libraries and evaluation utilities that integrate cleanly with TensorFlow ecosystems.
6 years of coding experience
University of California San DiegoHong Kong University of Science and Technology (HKUST)
Github Skills (8)
keras10
machine-learning10
privacy10
tensorflow10
python10
scikit-learn9
scikit9
testing7
Programming languages (2)
Github contributions (5)
tensorflow/privacy
Mar 2020 - Nov 2022
Library for training machine learning models with privacy for training data
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 43 commits, 6 PRs in 2 years 8 months
Contributions summary:Shuang's commits primarily focus on developing and integrating membership inference attack (MIA) techniques within the `tensorflow/privacy` repository. Their work involves implementing training hooks and functions for both TensorFlow Estimator and Keras models to perform membership inference attacks. These contributions include integrating new attack methods and creating utilities for model evaluation and reporting results, ultimately aimed at assessing privacy risks.
machine-learning-trainingprivacydifferential-privacymachine-learningtraining
google-research/DP-FTRL
Apr 2021 - Dec 2021
DP-FTRL from "Practical and Private (Deep) Learning without Sampling or Shuffling" for centralized training.
Contributions:2 commits, 1 push, 1 branch in 8 months
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.