About Me

Hi, I am Zixuan (Eve) Yi. I am a second-year CS PhD student at University of Pennsylvania, advised by Ryan Marcus and Zack Ives. Before this, I earned my bachelor degree in CS from Tsinghua University.

You can find me at Twitter @yi_zixuan, linkedin or email me zixy@seas.upenn.edu.

I’m always happy to chat — feel free to drop me an email about research 😄 or anything else on your mind.

News

🇩🇪 I will be attending the SIGMOD’25 Conference at Berlin, Germany.

Research Interests

I am broadly interested in Machine Learning and Systems.

Publications

LimeQO: Low-Rank Learning for Offline Query Optimization.
SIGMOD’25
Zixuan Yi, Yao Tian, Zachary G. Ives, Ryan Marcus [code][paper][poster]

Regressions and large overhead time are the major concerns when machine learning based query optimizers are put into production system. We proposed the first “workload-level” query optimizer that optimize the whole workload all at once without any regressions. Leveraging the similarities between queries, we cast the problem as low-rank matrix completion - using this pure linear method gives us 100x overhead time gain! Try it out at LimeQO.

The Unreasonable Effectiveness of LLMs for Query Optimization.
ML4Systems@NeurIPS’24 (🔦Spotlight)
Peter Akioyamen, Zixuan Yi, Ryan Marcus [code][paper][talk]

We present LLMSteer, a novel approach that leverages LLMs to enhance SQL query optimization. By embedding raw SQL queries using LLMs and training a simple binary classifier on a small labeled dataset, the method predicts optimal query plan hints. Remarkably, this approach outperforms traditional heuristic-based systems and complex machine learning models that require deep integration with database internals.