Visiting Student - MCS - Lai, Zhisen - 1.8.26.

Argonne National Laboratory • Lemont, IL

1d

About The Position

This project focuses on using AI to improve the data quality of scientific lossy compressors through fine-grained, point-wise optimization, while jointly modeling region-level error and uncertainty distributions induced by lossy compression. Specifically, we will (1) investigate how to leverage AI/ML models to post-process and improve the reconstruction quality (e.g., PSNR and SSIM) of data compressed by state-of-the-art scientific lossy compressors (e.g., SZ3 and STZ), and (2) study how AI/ML models can be used to model the uncertainty (i.e., error distributions) introduced by these compressors, enabling quantitative characterization of compression-induced uncertainty and improved user awareness of its impact on scientific data. Both methods will be designed to respect the quantities of interest (QoIs) and regions of interest (ROIs) in the data to achieve higher efficiency, and will be integrated into existing scientific compression pipelines and evaluated on real-world HPC datasets (e.g., Nyx and WarpX).

Requirements

• The entirety of the appointment must be conducted within the United States.
• Must be 18 years or older at the time the appointment begins.
• Applicants must be: Currently enrolled in undergraduate or graduate studies at an accredited institution. Graduated from an accredited institution within the past 3 months; or Actively enrolled in a graduate program at an accredited institution.

Responsibilities

• investigate how to leverage AI/ML models to post-process and improve the reconstruction quality of data compressed by state-of-the-art scientific lossy compressors
• study how AI/ML models can be used to model the uncertainty introduced by these compressors, enabling quantitative characterization of compression-induced uncertainty and improved user awareness of its impact on scientific data
• design methods to respect the quantities of interest (QoIs) and regions of interest (ROIs) in the data to achieve higher efficiency
• integrate methods into existing scientific compression pipelines and evaluated on real-world HPC datasets

What This Job Offers

• Job Type: Part-time
• Career Level: Intern
• Education Level: No Education Listed
• Number of Employees: 1,001-5,000 employees