Lily Liu

PhD Candidate

email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts
email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts

Summary

🤩
Rockstar
Lily Liu is a Member of Technical Staff at OpenAI with nine years of experience building high-performance ML and systems software. She holds a PhD in Computer Science from UC Berkeley and a strong academic track record including a master's from Carnegie Mellon and dual bachelor's from Peking University. Lily has a proven focus on inference and performance engineering for large language models—her open-source work on vLLM includes FlashInfer integration, RoPE scaling support, and speculative decoding optimizations. Her background spans internships and research roles at AWS, Anyscale, Alibaba, Snowflake, and Google, plus production-focused systems work at CMU. Colleagues describe her as a pragmatic researcher who translates deep algorithmic understanding into lean, memory-efficient implementations. She combines rigorous academic training with hands-on tuning of attention mechanisms and decoding pipelines that materially improve LLM throughput.
code9 years of coding experience
github-logo-circle

Github Skills (8)

pytorch10
inference10
performance-optimization10
python10
llm10
roc9
cuda9
transformer9

Programming languages (4)

C++HTMLPythonCuda

Github contributions (5)

github-logo-circle
vllm-project/vllm

Jun 2023 - Apr 2025

A high-throughput and memory-efficient inference and serving engine for LLMs
Role in this project:
userML Engineer & Performance Engineer
Contributions:204 reviews, 71 PRs, 41 pushes in 1 year 9 months
Contributions summary:Lily primarily focused on optimizing the performance and correctness of the vLLM inference engine for LLMs. Their contributions included removing unnecessary code, fixing maximum sequence length limitations, and implementing FlashInfer for decoding. They also worked on supporting RoPE scaling and improving the ngram lookup performance for speculative decoding. The user demonstrated expertise in the inner workings of the vLLM project, especially within the attention mechanisms and speculative decoding.
amdcudadeepseekgpthpu
LiuXiaoxuanPKU/papers

Nov 2017 - May 2022

Contributions:3 pushes, 1 branch in 4 years 6 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.
Request Free Trial
Lily Liu - PhD Candidate