Overview & Focus

The One-U Responsible AI Infrastructure Special Interest Group (RAI SIG) brings together practitioners, engineers, researchers, and policymakers working at the intersection of infrastructure and responsible AI development. Our primary focus is to explore, design, define, and share regional needs and best practices in infrastructure strategies that support transparency, fairness, security, audibility, and accountability in AI systems at scale.

We aim to build a forum where key stakeholders can:

• Define infrastructure needs for supporting Utah’s training, research, industrial, and government AI ecosystem.
• Share implementation strategies from industry, academia, and open-source communities.
• Showcase tools and practices that enable ethical AI development across the full model lifecycle.
• Assemble a regional cadre of experts in the practical use of AI infrastructure.

Key Activities & Timeline

Presentation Highlights

Our monthly talk series showcases in-depth, practical perspectives on responsible AI infrastructure. The following are selected highlights: 

Next Talk:

Topic: AI in Business and Marketing

Speaker: Himanshu Mishra, David Eccles School of Business, University of Utah

Past Talks:

“Why HPC Matters for Quantum Right Now”

Speaker: Alexey Galda, Moderna
Focus: Quantum matters to HPC and AI infrastructure conversations now not because fault-tolerant hardware is around the corner, but because today's quantum research is already computationally serious enough to require serious classical infrastructure to make meaningful progress.
Summary, Slide Deck

“ARUP Labs and Pathology Foundation Models”

Speaker: Brendan O’Fallon, ARUP Labs
Focus: Brendan presented on the ARUP team’s work-in-progress on a foundation model for hematopathology. Hematopathology is the study of blood disorders, and often involves examination of bone marrow, peripheral blood, and other tissues with light microscopy.  Currently, most of the basic analyses are very manual and require pathologists to manually count cells, estimate stain percentages, and scan tissue for abnormal cells. My team is creating a large vision transformer model that (we hope) will have general capabilities for identifying abnormal cells, tissues, stains, and other features of interest.

“GPU accelerated Alpha Fold for Protein Interaction Screens”

Speaker: Justin Bosch, University of Utah Genetics
Focus: Justin presented on how his team is using AlphaFold to perform large-scale in-silico protein interaction screens at the University of Utah Center for High Performance Computing (CHPC). Topics included: the general importance of this approach and how his lab is using it to study hormones and human genetic variation and Justin also shared insights from his lab’s real-world experiences with CHPC’s computing infrastructure, including comparisons of AlphaFold on 17 different GPUs, how protein size affects GPU choice, struggles to take full advantage of GPU resources at the CHPC, and rationale behind future GPU purchasing decisions.

Additional Presentations by:

• Martin Cuma, Center for High Performance Computing, University of Utah;
• Vivek Srikumar, Kahlert School of Computing, University of Utah;
• Jim Stewart, Utah Education Network;
  
• And more!

Highlights and Opportunities

• Presentations have been well-attended and draw participation from cross-functional roles: ML engineers, policy teams, infrastructure architects, and open-source contributors.
• Creation of a “Use Case” description sight to give the ecosystem an idea of what is happening in Utah that would gain by having significant infrastructure available

Next Steps & How to Get Involved

• We welcome new members interested in shaping responsible AI through infrastructure innovation.
• The SIG is actively recruiting contributors for our white paper working group and upcoming workshop sessions.
• We’re also happy to support outreach to previous speakers for content if you’d like to feature their talks more broadly.

Contact: Greg Jones, Sara McLaughlin Wynn

Get Involved

Upcoming Events and Engagements

AI Infrastructure SIG meetings are typically held via Zoom from 9:00-10:00 on the first Friday of the month. View the calendar for meeting dates.

Sign up to receive the monthly calendar invitation by editing your name in the subject line and sending an email message through this link.