Xiangyang Ju

Ph.D in Physics

I build AI systems that connect scientific intent with data, software, and large-scale computing. At Berkeley Lab, I develop machine-learning methods and inference infrastructure for particle physics.

I am a Computing System Engineer in the Scientific Data Division at Lawrence Berkeley National Laboratory and a member of the ATLAS Collaboration at the Large Hadron Collider. My research asks how AI and high-performance computing can reduce the effort between a scientific idea and a reproducible result.

Explore my research Get in touch

Scalable ML infrastructure

Inference as a Service

Bringing GPU-accelerated models to physics applications through shared inference services and integration with HPC facilities.

Selected work

  • Tracking as a service: separating track reconstruction from the client application so accelerator resources can be shared. 2025 paper.
  • Learning to find tracks: TrackSorter explores Transformer-based sorting for particle track finding. Paper.
  • Generative simulation: incorporating particle flavor into deep learning models for hadronization. Paper.

Selected publications and full list · Talks and posters

Working together

I enjoy working with students and collaborators across physics, machine learning, and scientific computing. I welcome conversations about agentic workflows, scientific representations, and scalable inference. Email me to discuss a research connection, or explore my mentoring record.

Education

Ph.D. in Physics, University of Wisconsin–Madison, March 2018

Dissertation: Observation of a Standard Model Higgs boson and search for additional heavy scalars in the $\ell\ell\ell\ell$ final state with the ATLAS detector.

Advisor: Prof. Sau Lan Wu

B.S., Nanjing University, July 2009

Thesis: Measurement of the top mass using the ATLAS detector.

Advisor: Prof. Shenjian Chen