Jerry Liu
Software Engineer - Ads at Google
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolJerry Liu is a software engineer specializing in ads at Google with 11 years of experience building production systems and ML-driven features across YouTube and Google. Trained in biomedical engineering at the University of Waterloo with an exchange at KAIST, he blends rigorous quantitative foundations with practical product instincts from internships at Apple and platform work on Android. His open-source contributions to high-profile repositories like tensorflow/models and Google’s uncertainty-baselines focus on Gaussian processes, Monte Carlo Dropout, and spectral-normalized layers—efforts that strengthen model uncertainty quantification and robustness. Based in New York, he brings cross-domain experience from live-streaming systems to ad infrastructure, delivering measurable improvements in reliability and performance. Notably, his Github tagline, “Statistical Intelligence for Human Good,” reflects a consistent emphasis on principled probabilistic modeling to make ML systems safer and more trustworthy.
11 years of coding experience
5 years of employment as a software developer
Exchange Term School of Electrical Engineering, Exchange Term School of Electrical Engineering at Korea Advanced Institute of Science and TechnologyMaCS Program Ontario Secondary School Diploma, MaCS Program Ontario Secondary School Diploma at William Lyon Mackenzie Collegiate InstituteBachelor of Applied Science (B.A.Sc.) Biomedical Engineering, Bachelor of Applied Science (B.A.Sc.) Biomedical Engineering at University of Waterloo
English, Chinese, French
Github Skills (11)
neural-network10
machine-learning10
nlp10
uncertainty-quantification10
deep-learning10
tensorflow10
python10
bert10
gaussian-processes10
keras9
pytorch4
Programming languages (4)
Github contributions (5)
google/uncertainty-baselines
Aug 2020 - Nov 2022
High-quality implementations of standard and SOTA methods on a variety of tasks.
Role in this project:
ML Engineer
ML EngineerContributions:7 reviews, 128 commits, 2 PRs in 2 years 4 months
Contributions summary:Jerry made several commits focused on integrating and extending BERT-based models within the "uncertainty-baselines" framework. Key contributions include moving shared utility functions to a dedicated file, ensuring consistent batch sizes during training and testing, adding an AUC metric for OOD detection to a CLINC baseline, integrating Monte Carlo Dropout to ImageNet ResNet50, adding both MC Dropout and SNGP methods to ImageNet baseline, and implementing a spectral-normalized einsum dense layer. These changes indicate a focus on uncertainty quantification and improving model robustness.
implementationsstatisticsdata-sciencedeep-learningneural-networks
tensorflow/models
Mar 2021 - Apr 2021
Models and examples built with TensorFlow
Role in this project:ML Engineer
ML EngineerContributions:22 commits, 3 comments in 1 month
Contributions summary:Jerry's contributions center around implementing and integrating Gaussian Process (GP) layers within the TensorFlow models framework. Their work involved creating a `RandomFeatureGaussianProcess` layer and a `GaussianProcessClassificationHead` for classification tasks. They also added unit tests for the implemented GP layers, demonstrating a focus on model uncertainty quantification and improving model performance through the use of Gaussian processes.
deep-learningtensorflow
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.