Peter Liu
Co-Founder at Compound AI
California, United States
Summary
🤩
Rockstar🎓
Top SchoolPeter Liu is a Co-Founder and AI researcher with 12 years of experience building and scaling language-model-driven products from Google Brain and DeepMind to startups in California. He led foundational LM work at Google (including early Transformer decoder-only models and T5-era research) and shipped applied ML systems at scale such as Gmail's first large neural spam filter. At DeepMind he led Science of Scaling and contributed to models like Gemma 2 and tooling such as nanodo, then transitioned to founding startups focused on bringing advanced AI to finance. An active open-source contributor, he improved ROUGE evaluation and added key summarization and translation datasets to tensorflow/datasets, helping reproducible evaluation in the community. Known for bridging deep research with productization, he combines rigorous theoretical background (CompSci, pure math, ML) with a pragmatic track record of deploying generative AI features.
12 years of coding experience
13 years of employment as a software developer
Computer Science Pure Mathematics Machine Learning, Computer Science Pure Mathematics Machine Learning at University of Toronto
French
Github Skills (11)
machine-learning10
nlp10
tensorflow10
python10
user-data9
data-model9
datasets9
data-set9
ai9
testing8
jax7
Programming languages (5)
Github contributions (5)
google-research/google-research
Sep 2019 - Jul 2022
Google Research
Role in this project:
ML Engineer
ML EngineerContributions:19 commits, 7 comments in 2 years 10 months
Contributions summary:Peter primarily contributed to the `rouge` directory, implementing and testing ROUGE metrics. Their work involved adding support for `rougeLsum` and fixing related edge cases. They also updated the project to support Python 3 and made improvements to the codebase, including removing recursion for large inputs.
googlemachine-learningai
tensorflow/datasets
Nov 2019 - Sep 2022
TFDS is a collection of datasets ready to use with TensorFlow, Jax, ...
Role in this project:ML Engineer
ML EngineerContributions:6 commits, 2 comments in 2 years 10 months
Contributions summary:Peter contributed to the tensorflow/datasets repository by implementing and modifying datasets related to text summarization and machine translation. Specifically, the user added the QuALITY, scrolls, and MTNT datasets, and enhanced the CNN/DailyMail dataset by adding publisher and ID features. These changes suggest a focus on improving the dataset's functionalities for machine learning tasks.
datanumpydeep-learningdatasetmachine-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.