Mingrui Liu
Assistant Professor at George Mason University
Virginia, United States
Summary
🤩
Rockstar🎓
Top SchoolMenrui Mr is a research scientist and deep learning engineer with eight years of experience, currently contributing to DAMO Academy at Alibaba Group. He specializes in multimodal models and image generation, with notable work on the official OFA (ICML 2022) repository where he integrated CLIP-based evaluation, added caption-inference tooling, and hardened the image generation pipeline. He also extended and refactored the Chinese-CLIP codebase to support large ViT models (ViT-L/H) and fixed production-facing bugs like RN50 loading issues. His contributions reveal a practical researcher’s mindset—bridging cutting-edge papers with reproducible tooling, dataset preprocessing, and inference-ready APIs so models are usable beyond experiments.
8 years of coding experience
Doctor of Philosophy (Ph.D.), Computer Science, Doctor of Philosophy (Ph.D.), Computer Science at University of IowaMathematics, Mathematics at Shanghai Jiao Tong University
Chinese, English
Github Skills (17)
pytorch10
python10
multimodal10
image-annotation10
captions10
multi-modal10
pretrained-model10
deep-learning10
image-generation10
clip10
computer-vision10
image-text10
caption10
machine-learning9
nlp8
Programming languages (3)
Github contributions (5)
OFA-Sys/Chinese-CLIP
Jul 2022 - Jan 2023
Chinese version of CLIP which achieves Chinese cross-modal retrieval and representation generation.
Role in this project:
ML Engineer
ML EngineerContributions:14 commits, 3 PRs, 5 pushes in 6 months
Contributions summary:Mingrui primarily focused on modifying and extending the `cn_clip` library, a Chinese version of CLIP for cross-modal retrieval. Their contributions included refactoring code and adding API usage examples, and updating the code to support new models like ViT-L-14, ViT-H-14, and RN50. Furthermore, the user fixed a bug related to loading the RN50 model and updated the setup file.
chineseretrievalrepresentationtransformerscontrastive-loss
OFA-Sys/OFA
Feb 2022 - Jun 2022
Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
Role in this project:ML Engineer
ML EngineerContributions:28 commits, 3 PRs, 14 pushes in 3 months
Contributions summary:Mingrui primarily contributed to the image generation aspects of the OFA repository, focusing on the integration and use of a CLIP model for image evaluation and generation. They added a notebook for caption inference and uploaded a text-to-image inference shell. The user also made several changes to datasets and scripts, as well as fixed bugs related to the image generation pipeline, including the data preprocessing steps and model configurations. Their work appears focused on building and refining the image generation capabilities.
chinesesequencevisual-question-answeringframework-learningicml
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.