Frederick Liu
Senior Staff Software Engineer at Google DeepMind
United States
Summary
🤩
Rockstar🎓
Top SchoolFrederick Liu is a Senior Staff Software Engineer with a decade of experience building production-grade ML systems and libraries, now at Google DeepMind after six years leading core modeling work at Google. He designed constrained decoding, parameter-efficient fine-tuning, structured prediction, and interpretability tooling used across large language and vision models (Gemini, PaLM, T5), and has published research at NeurIPS, ACL and COLM. His contributions to the high-profile tensorflow/models repo include practical NLP training and evaluation improvements—adding an open-source translation task and BLEU computation—to tighten model training and metrics pipelines. Earlier work spans on-device federated learning with differential privacy at Snap and multimodal machine translation research from Carnegie Mellon, reflecting a blend of privacy-conscious engineering and academic rigor. Frederick pairs deep research insight with hands-on delivery, often transforming experimental ideas into reliable libraries that power both research and products. Colleagues describe him as an explorer who consistently ships robust, production-ready ML infrastructure.
10 years of coding experience
9 years of employment as a software developer
Bachelor's degree Computer Science, Bachelor's degree Computer Science at National Taiwan UniversityMaster's degree at Language Technology Institute Computer Science, Master's degree at Language Technology Institute Computer Science at Carnegie Mellon University
English, Chinese, Japanese
Github Skills (9)
sentencepiece10
machine-learning10
transformer10
nlp10
tensorflow10
bleu10
trainings10
evaluation10
modeling10
Programming languages (6)
Github contributions (5)
tensorflow/models
Dec 2020 - Oct 2022
Models and examples built with TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:20 reviews, 171 commits, 8 PRs in 1 year 10 months
Contributions summary:Frederick made internal changes and added features related to training and evaluating sentencepiece models for NLP tasks. Their contributions included modifying code for masked language modeling, specifically avoiding computations when inputs lack masked positions. They also introduced an open-source translation task and updated the code to compute BLEU scores, and moved relevant code to metrics. These changes suggest a focus on model training and evaluation pipelines within the TensorFlow models repository.
deep-learningtensorflow
frederick0329/TracIn
Jun 2020 - Feb 2022
Implementation of Estimating Training Data Influence by Tracing Gradient Descent (NeurIPS 2020)
Contributions:37 commits, 1 PR, 34 pushes in 1 year 7 months
pytorchinfluenceneurips-2020deep-learningtracing
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.