Geyang (Albert) Xu

I am Albert (Geyang) Xu, a Ph.D. student in Computer Science at UC Riverside, advised by Prof. Qian Zhang. My research focuses on software engineering for machine learning, especially the testing and reliability of deep learning compilers and AI systems.

I hold a Master’s degree in Computer Science from UC San Diego (GPA: 3.87/4.0) and a Bachelor’s degree from the University of Liverpool (GPA: 3.93/4.0).

Previously, I was a Manufacturing Data Analyst at Fuyao Glass America, building Python/SQL data pipelines and production monitoring jobs, and a Back-end Engineer at 4Pexonic. Before that, I was a Research Engineer at the UCSD Halıcıoğlu Data Science Institute (HDSI), where I co-authored research on stress-testing ML pipelines (VLDB 2026).

My research interests include:

• Deep Learning Compiler Testing: Systematically testing deep learning compilers for correctness and robustness bugs.
• Reliability of AI Systems: Making ML and AI systems behave predictably, and explaining why they fail.
• Data-Centric Robustness: Stress-testing ML pipelines with adversarial data corruption.

Research Philosophy

At Fuyao, I spent my days building analytics layers, designing experiments, and monitoring production to make our data systems transparent, stable, and traceable. At 4Pexonic, I witnessed the other side of the data lifecycle—how large-scale bibliographic data is collected, cached, and aggregated to produce metrics like “research impact” and “disciplinary trends.” There, technical decisions like caching strategies and latency handling directly shaped how the academic ecosystem was perceived.

These experiences reinforced a common theme: engineering roles are often about executing within existing rules and standards. I can optimize a coating line’s data pipeline or tune an analytics platform for low latency, but I have little influence over how we evaluate and govern these systems in the first place.

What truly fascinates me is not just maintaining stability, but defining the general frameworks for reliability and governance in the next generation of ML and data pipelines. How should we describe and inject data errors? How do we perform systematic stress testing on complex pipelines? How do we translate these results into evidence that engineers, stakeholders, and regulators can trust?

These questions go far beyond individual engineering practice; they require systematic research and community consensus. Pursuing a Ph.D. is my path from executing existing standards to helping define them.