Yonglong Tian
Member Of Technical Staff at OpenAI
Cambridge, Massachusetts, United States
Summary
🤩
Rockstar🎓
Top SchoolYonglong Tian is a research-focused machine learning engineer and scientist with eight years of experience across top AI labs, currently a Member of Technical Staff at OpenAI after roles at Google Research and DeepMind. He holds advanced degrees including a Ph.D. from MIT and has a strong track record in contrastive learning and knowledge distillation, contributing notable open-source implementations such as Contrastive Multiview Coding and RepDistiller. Yonglong’s work blends deep learning research with practical model engineering—fixing pre-trained model integration for detection tasks and adapting ResNet backbones for contrastive frameworks. Based in Cambridge, MA, he brings research rigor and production-minded code contributions, and his trajectory from internships at NVIDIA and Google to senior research roles reflects a fast, research-driven climb in AI. An understated strength is his penchant for reworking model architectures and distributed-training details that enable academic ideas to scale in real projects.
8 years of coding experience
2 years of employment as a software developer
Ph.D Computer Science, Ph.D Computer Science at Massachusetts Institute of TechnologyB.Eng., B.Eng. at Tsinghua UniversityThe Chinese University of Hong Kong (CUHK)
English, Chinese
Github Skills (13)
neural-network10
computer-vision10
pytorch10
machine-learning10
distill10
deep-learning10
resnet10
python10
dis10
image-classification9
modeling8
trainings8
tensorboard6
Programming languages (2)
Github contributions (5)
HobbitLong/CMC
Jun 2019 - Nov 2020
[ECCV 2020] "Contrastive Multiview Coding", also contains implementations for MoCo and InstDis
Role in this project:
ML Engineer
ML EngineerContributions:48 commits, 3 PRs, 45 pushes in 1 year 5 months
Contributions summary:Yonglong implemented and modified ResNet models within the repository, including the addition of various ResNet architectures and the modification of existing ones. Their work also involved the removal of a `.compute_feat` function, which may have been related to distributed training considerations. Furthermore, they made changes to support ResNet backbones, suggesting an effort to integrate these models into the project's architecture and potentially leverage their features for contrastive learning or related tasks.
pytorchimplementationsmultiviewunsupervised-learningeccv
HobbitLong/PyContrast
May 2020 - Mar 2021
PyTorch implementation of Contrastive Learning methods
Role in this project:ML Engineer
ML EngineerContributions:1 review, 46 commits, 7 PRs in 10 months
Contributions summary:Yonglong primarily contributed to the `pycontrast` repository, which focuses on Contrastive Learning methods implemented in PyTorch. Their work involved fixing loading issues related to pre-trained models, indicating they were working with the model's architecture and weights. They also implemented and updated functionality related to the conversion of pre-trained models for use in detection tasks. These commits demonstrate a focus on model integration and modification within a computer vision project.
pytorchunsupervised-learningdeep-learningcontrastive-learningcontrastive
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.