Ying Sheng
Co-Founder (CEO) at RadixArk
Palo Alto, California, United States
Summary
🤩
Rockstar🎓
Top SchoolYing Sheng is a machine learning systems engineer and founder with nine years of experience building high-performance ML infrastructure and formally verified software. Currently CEO of RadixArk in Palo Alto, she previously contributed to production LLM serving and optimization at xAI and Databricks Mosaic Research, and has deep research roots from a Stanford PhD and internships at Meta and Microsoft Research. Her open-source work spans running large language models efficiently on a single GPU (FlexLLMGen), serving frameworks for LLMs/VLMs (SGLang), and core improvements to the cvc5 theorem prover, reflecting a rare blend of systems performance engineering and automated reasoning. Comfortable moving between research and production, she has optimized GPU memory/access patterns and extended SMT solver architectures—skills that surface in both shipping scalable LLM services and advancing formal verification.
9 years of coding experience
2 years of employment as a software developer
Bachalor's Degree Computer Science, Bachalor's Degree Computer Science at Shanghai Jiao Tong UniversityMaster of Science (M.S.) Computer Science, Master of Science (M.S.) Computer Science at Columbia UniversityDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Stanford University
Chinese, English
Github Skills (31)
algorithm10
algorithms10
pytorch10
c-language10
python10
architecture10
solver10
gpu-programming10
large-language-models10
data-structure10
sat-solver10
deeplearning-ai10
deep-learning10
gpu10
optimisation10
Programming languages (8)
Github contributions (5)
FMInference/FlexLLMGen
Feb 2023 - Mar 2023
Running large language models on a single GPU for throughput-oriented scenarios.
Role in this project:
ML Engineer
ML EngineerContributions:4 reviews, 61 commits, 47 PRs in 20 days
Contributions summary:Ying contributed to the `flexllmgen` repository, which focuses on running large language models on a single GPU. Their commits primarily involve integrating and modifying DeepSpeed components for memory access optimization, enhancing performance. The user also worked on updating and improving the chatbot and completion applications within the project. Furthermore, they added a HELM-based text summarization example, showcasing contributions to the project's functionality and practical applications.
deep-learninggpt-3high-throughputlarge-language-modelsmachine-learning
sgl-project/sglang
Jan 2024 - Mar 2025
SGLang is a fast serving framework for large language models and vision language models.
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:8 releases, 204 reviews, 389 PRs in 1 year 2 months
Contributions summary:Ying's contributions primarily revolve around addressing and resolving issues related to GPU utilization, specifically for T4 GPUs. The code changes suggest involvement in optimizing the framework for large language models and vision language models, as indicated by the file modifications to flash attention. The user's involvement with the sglang framework indicates contributions in developing serving framework for LLMs. Additional changes in model rpc server suggest contributions in building serving infrastructure.
cudadeepseekdeepseek-llmdeepseek-r1deepseek-r1-zero
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.