Liangzhe Yuan
Staff Research Scientist at Google DeepMind
Los Angeles, California, United States
Summary
👤
Senior🎓
Top SchoolLiangzhe Yuan is a Staff Research Scientist based in Los Angeles with eight years of experience tackling perception and general AI problems across industry-leading teams. At Google and now Google DeepMind he has driven multimodal and video foundation model work, including contributions to Gemini Video, Astra Agent, and multimodal content co-creation. His published engineering work spans TensorFlow and Google Research—touching LSTM-based object detection, video classification, ViT positional embedding tricks, and pose-estimation pipelines—demonstrating a blend of research rigor and production-level code. Trained in robotics (UPenn MS) with a mechanical engineering undergraduate background, he brings hands-on systems intuition from drone VIO/obstacle avoidance to large-scale self-supervised representation learning. He is a pragmatic open-source contributor who nudges academic ideas into robust, testable implementations and seems to approach problems with a "learning in everything" mindset.
8 years of coding experience
5 years of employment as a software developer
Bachelor of Engineering, Mechanical Engineering, Bachelor of Engineering, Mechanical Engineering at Shanghai Jiao Tong UniversityMaster of Science in Engineering, Robotics, Master of Science in Engineering, Robotics at University of Pennsylvania
Chinese, English, German
Github Skills (13)
object-detection10
keras10
computer-vision10
machine-learning10
tensorflow10
data-augmentation10
python10
pose-estimation10
classification10
classify10
ai9
lstm9
vi9
Programming languages (2)
Github contributions (5)
tensorflow/models
Mar 2020 - Nov 2022
Models and examples built with TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:72 commits, 2 PRs in 2 years 8 months
Contributions summary:Liangzhe contributed to the TensorFlow models repository by modifying and testing components related to machine learning models, specifically focusing on the LSTM-based object detection and video classification tasks. Their commits included code changes to the lstm_ssd_mobilenet_v1_feature_extractor, fixes in the lstm_cells, as well as the refactoring of resnet_3d. Furthermore, the user supported the use of separable_conv in CenterNet and worked on the ViT model, including dynamic positional embedding interpolation.
deep-learningtensorflow
google-research/google-research
Dec 2020 - Feb 2021
Google Research
Role in this project:ML Engineer
ML EngineerContributions:6 commits in 2 months
Contributions summary:Liangzhe primarily contributed to the POEM (Pose Estimation and Modeling) project within Google Research. Their work involved modifying and refactoring code related to keypoint processing, camera augmentation, and model functionalities. They also made changes to testing files, suggesting a focus on ensuring the robustness and accuracy of the pose estimation pipeline. Key contributions include updating the `keypoint_utils` module to support sequential camera augmentation and modifying `model.py`, as well as internal changes to various test files.
googlemachine-learningai
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.