Chang Su
Principal Software Engineer at Oracle
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolChang Su is a Principal Software Engineer based in Seattle with a decade of experience building scalable ML-infused backend systems and production-grade serving engines. His career spans high-impact roles at Oracle, Meta, and X (Intrinsics), where he shipped full-stack and robotics ML infrastructure including recurring training pipelines and sub-millimeter control applications. He is an active open-source contributor to prominent ML-serving projects—helping harden vllm and SGLang for LLM and vision-model inference—demonstrating deep expertise in model integration, embedding APIs, and request validation. Past research roles at USC and work on reproducible reinforcement learning show a strong grounding in both academic ML and industrial engineering. Known for pragmatic bug fixes that improve reliability at scale, he combines hands-on backend engineering with ML systems design to close the gap between research models and production deployment.
10 years of coding experience
11 years of employment as a software developer
Master's degree Computer Science, Master's degree Computer Science at University of Southern CaliforniaBachelor of Engineering (B.Eng.) Software Engineering, Bachelor of Engineering (B.Eng.) Software Engineering at Beijing University of Posts and Telecommunications
English
Github Skills (24)
algorithm10
algorithms10
pytorch10
rep10
debug10
python10
testing10
machine-learning10
lm10
inference10
reinforcement-learning10
llm10
tensorflow10
repr10
transformer10
Programming languages (2)
Github contributions (5)
rlworkgroup/garage
Jun 2018 - Jun 2019
A toolkit for reproducible reinforcement learning research.
Role in this project:
ML Engineer
ML EngineerContributions:36 commits, 81 PRs, 345 pushes in 11 months
Contributions summary:Chang primarily contributed to the development of reinforcement learning algorithms within the garage repository, focused on reproducible research. Their work involved implementing dynamics randomization techniques for MuJoCo environments, specifically involving data structures and wrapper classes. They also made significant changes to tensorboard integration, including fixing TF device issues, adding name scopes for better graph organization, and updating examples. Furthermore, the user added exploration strategies, such as Ornstein-Uhlenbeck, and other core components for DDPG, including a replay buffer.
pytorchreinforcement-learningreproduciblereproducibilityrl-algorithms
vllm-project/vllm
Jan 2024 - Feb 2025
A high-throughput and memory-efficient inference and serving engine for LLMs
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:42 reviews, 13 PRs, 100 comments in 1 year 1 month
Contributions summary:Chang contributed to bug fixes and improvements within the VLLM project, specifically addressing issues in benchmark scripts and embedding API implementations. They added support for new models, such as e5-mistral-7b-instruct, and modified the core model runner for embedding models. The user also addressed a critical issue related to temperature settings and empty outputs, and fixed tests. These contributions indicate a strong focus on model integration, API functionality, and the overall reliability of the LLM serving engine.
amdcudadeepseekgpthpu
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.