Volume 8 - August 28, 2026¶

Scaling to a (Free) National Resource

As artificial intelligence transforms medicine, the need for transparent, trustworthy models is more critical than ever. At the University of Arizona, fifth-year computer science PhD student Saiful Islam is tackling this challenge head-on by examining the inner workings of large language models (LLMs).

Saiful’s research focuses on Explainable AI (XAI) within healthcare and bioinformatics. Modern LLMs often operate as "black boxes," providing answers without revealing their underlying reasoning. While high accuracy may work for everyday tasks, high-stakes clinical settings demand clear accountability. As Saiful notes, high accuracy alone isn't enough when human lives are involved—explainability is essential for medical professionals to safely trust AI diagnostics.

However, training and interpreting massive language models requires immense computational power, and local university servers were quickly stretched thin. To overcome these bottlenecks, Saiful turned to the NSF ACCESS community and Jetstream, an advanced cloud infrastructure designed for high-performance computing.

With support from local facilitators, Saiful initially gained access to four NVIDIA H100 GPUs, giving his team the raw power needed for demanding NLP training. He soon discovered that PhD students can apply directly as Principal Investigators for their own ACCESS allocations. After submitting a proposal, his allocation was approved in just one day, allowing him to expand his work to national supercomputing resources like Delta AI at NCSA.

This computing power enabled Saiful to complete a major research paper now prepared for journal submission. His journey demonstrates how ACCESS and Jetstream bridge the gap between ambitious ideas and real breakthroughs by offering frictionless access to GPU clusters and expert support.

For researchers facing compute constraints, ACCESS and Jetstream provide a streamlined pathway to supercomputing power. Explore available allocations today to accelerate your research.

Welcoming New Wildcats

The Research & Discovery Technologies team attended Graduate Orientation on Aug. 19 to let new master's and doctoral students know about campus computing resources available to them. From HPC compute time at no cost to supercomputing consultation and workshops to expert visualization support to computing resources for researchers working with regulated data, R&DT has services to give the next generation of researchers a strong start to the Fall 2026 Semester.