Juncheng Liu
AI Research Scientist at Meta
Beijing, Singapore
Summary
🤩
Rockstar🎓
Top SchoolJuncheng Liu is an AI research scientist with 11 years of experience bridging deep learning research and production systems, currently working on foundation models for time series at Meta after research roles at Microsoft and Salesforce. He holds a PhD in Computer Science from NUS with a strong background in graph representation learning, GNNs, and knowledge graphs, and has collaborated closely with several leading academics. His work at Salesforce produced Moirai 2.0 and a MoE variant for time series, plus Gift-Eval, a large benchmarking suite covering 177 million data points, and he has led predictive autoscaling across 600+ Kubernetes cells. A pragmatic engineer as well as researcher, he has contributed CUDA kernel optimizations and model parallelization to the OneFlow deep learning framework. Based in Beijing and Singapore-educated, he combines systems-level performance tuning with cutting-edge model design to scale AI from kernels to cloud services.
11 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering - BE, Computer Science and Technology, Bachelor of Engineering - BE, Computer Science and Technology at Sichuan UniversityXiamen No.1 High School of FujianDoctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at National University of Singapore
Chinese, English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (19)
c-language10
algebra10
deep-q-learning10
kernel10
numerics10
deep-learning10
parallel-computing10
numerical-methods10
cuda10
numerical-analysis10
cprogramming-language10
numerical-computing10
operation9
machine-learning9
tensorrt9
Programming languages (8)
Github contributions (5)
Oneflow-Inc/oneflow
Aug 2018 - Jan 2023
OneFlow is a deep learning framework designed to be user-friendly, scalable and efficient.
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:1375 reviews, 1582 commits, 697 PRs in 4 years 5 months
Contributions summary:Juncheng contributed to the OneFlow deep learning framework by implementing and optimizing CUDA kernels. Their work involved model parallelization using gather/unsorted_segment_sum, performance improvements to convolution operations, and enhancements to the fused batch normalization and ReLU operations. They also focused on addressing CUDA compatibility and incorporating the latest features in the library's codebase.
caffe2tensorflowframework-learningchainerdeep-learning-framework
Oneflow-Inc/diffusion-benchmark
Dec 2022 - Feb 2023
Contributions:18 commits, 25 PRs, 19 pushes in 2 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.