Xing Chen
Tempe, Arizona, United States
Summary
🤩
Rockstar🎓
Top SchoolXing Chen is the Technical Team Lead of Pretrain Infrastructure at StepFun, specializing in building and optimizing training stacks for billion-to-trillion-parameter LLMs and diffusion models across language, image, and video modalities. With a PhD from Arizona State University and three years of hands-on industry experience, he has led model-system co-design that doubled MFU and delivered a zero-overhead monitoring system tracking million-scale metrics per iteration to sustain over 98% effective training utilization. His work spans pretraining, midtraining, and SFT for Step-2 through Step-Video, and emphasizes long-context capabilities, fine-grained architectures, and large-scale multimodal training. Previously he contributed to distributed ML frameworks at ByteDance and developed dense/MoE training pipelines for thousands of GPUs as a senior engineer, bringing both research rigor and production-grade engineering to infrastructure. He is actively exploring next-generation AI infrastructure and model-system innovations that bridge academic advances with production reliability.
3 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy - PhD, Electrical and Electronics Engineering, 4.0, Doctor of Philosophy - PhD, Electrical and Electronics Engineering, 4.0 at Arizona State UniversityMaster's degree, Communication Engineering, Master's degree, Communication Engineering at Harbin Institute of Technology
Github Skills (14)
occupancy7
nvidia7
sdk5
deep-learning5
tensorrt5
machine-learning4
optimizer4
inference4
latency4
throughput4
tensorflow3
low-latency3
high-performance3
gpu-acceleration2
Programming languages (2)
Github contributions (5)
ccchen6/OccDepth
Mar 2023 - Mar 2023
Maybe the first academic open work on stereo 3D SSC method with vision-only input.
Contributions:2 pushes in 1 day
stepfun-ai/Step-Audio
Feb 2025 - Feb 2025
Contributions:8 PRs, 11 pushes, 3 comments in 9 days
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.