Zirui's Homepage

About me

I’m Zirui “Ray” Liu, a Ph.D. student from the Department of Computer Science at Rice University, co-advised by Dr. Xia “Ben” Hu and Prof. Vladimir Braverman.

My research interests lie in large-scale machine learning. I develop randomized algorithm and build systems to support them, aiming to scale-up and/or accerlate machine learning models. Some examples include randomized matrix mulplication, randomized matrix sparsification, stochastic quantization.

Recently, I’ve been exploring ways to train, fine-tune, and deploy Large Language Models (LLMs) on commodity hardware, especially making them more accessible to academics.

Email: zl105 at rice dot edu. Here is my CV.

I’m graduating in 2024 and am actively seeking a tenure-track faculty position.

Education

• Ph.D. in Computer Science, 2021 - now.
  Rice University, TX, USA

• Ph.D. in Computer Science, 2019 - 2021 (Transfer Out). Texas A&M University, TX, USA

• M.Eng. in Electrical Engineering and Automation, 2016 - 2018.
  Harbin Insuitute of Technology, Harbin, China

• B.Eng. in Electrical Engineering and Automation, 2012 - 2016.
  Harbin Insuitute of Technology, Harbin, China

News

• Our new work kivi shows that 2-bit is enough for representing KV Cache! It can improve inference throughput by 3X and provide direct support for long context. Please check our code to have a try.

• Our Memory-Efficient LLM fine-tuning work is covered by Rice CS New.

• Three paper accepted to Neurips.

• Two paper accepted to TMLR.

• Twos papers accepted to ICML 2023.

• One paper accepted to MLSys 2023.

• Two papers accepted to Neurips 2022, DreamShard and GNN Benchmark (Benchmark track).

Publications

Please refer to publications or Google Scholar.