Cosmo Du
Research Scientist at Meta
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolCosmo Du is an AI research scientist with 14 years of experience, currently at Meta TBD Lab after leading research teams as Principal Scientist & Director at Google DeepMind. He holds a PhD in Computer Science from Zhejiang University and has a strong track record in deep learning infrastructure and model development. His open-source contributions include improving MXNet’s ResNet implementations, multi-GPU training stability, and memory-leak fixes—work that boosted performance and reliability in a widely used deep learning framework. Based in Mountain View, he combines production-minded engineering with research rigor, bridging algorithmic innovation and systems-level robustness. An early project on video synopsis during his NUS internship hints at a long-standing interest in applying vision research to real-world surveillance and efficiency problems.
14 years of coding experience
8 years of employment as a software developer
Bachelor’s Degree Computer Science, Bachelor’s Degree Computer Science at East China University of Science and TechnologyDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Zhejiang University
Chinese, English
Github Skills (14)
mxnet10
computer-vision10
machine-learning10
deep-learning10
resnet10
python10
multi-gpu10
multiple-gpu10
image-classification10
debug9
algorithms8
data-structures8
algorithm8
data-structure8
Programming languages (2)
Github contributions (5)
apache/mxnet
May 2016 - Feb 2017
Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
Role in this project:
ML Engineer
ML EngineerContributions:7 commits, 7 PRs, 8 comments in 9 months
Contributions summary:Cosmo contributed to the implementation and refinement of ResNet models for image classification, specifically focusing on the CIFAR10 dataset. Their work included reproducing existing architectures, addressing multi-GPU training issues, and optimizing the training process. The user also fixed memory leaks and other bugs within the MXNet framework, demonstrating a strong understanding of the underlying infrastructure. These contributions enhanced the performance and stability of the MXNet deep learning framework.
pythonschedulerdataflowmutationdata-science
Answeror/RPCD
Dec 2011 - Jun 2012
Contributions:18 commits in 5 months
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.