Schedule

Jordan Ballroom

Jordan Ballroom

Join us to hear 1- minute lightning talks from our student poster presentations. 

Location- Jordan Ballroom
Bryan Smith, Acting Chief Technology Officer for Nuclear Science & Technology at the Idaho National Lab

Accelerating America’s Nuclear Future: Building Advanced Nuclear Infrastructure at Speed and Scale


The United States has entered a transformative period in nuclear energy development, driven by unprecedented load growth from data centers and AI infrastructure, coupled with the most ambitious federal nuclear directives in decades. This presentation examines the convergence of technology demonstration, industrial partnership, and policy acceleration that is reshaping America’s nuclear landscape. Drawing from ongoing work at Idaho National Laboratory, we’ll explore the reactor and fuel cycle pilot projects progressing from concept to concrete deployment, including INL’s role as the nation’s premier testbed for advanced reactor technologies. The presentation will detail emerging frameworks for industry collaboration that are enabling diverse off-takers, from hyperscale data centers to military installations, to partner in deploying next-generation nuclear systems. Finally, we’ll assess progress against the aggressive timelines established by last year’s landmark nuclear Executive Orders, which call for demonstrating multiple advanced reactor designs and achieving significant new nuclear capacity by decade’s end on the path to quadrupling American nuclear capacity by 2050.

Jordan  Ballroom- Room D

DLI Workshop - Data Parallelism: How to Train Deep Learning Models on Multiple GPUs
Presenter- Daniel Howard

With support of the Deep Learning Institute from NVIDIA, a training workshop is offered to all RMACC attendees. Attendees will also be provided information on how to become certified instructors, such as with community support from Cyberinfrastructure Community-wide Mentorship Network (CCMNet), in order to offer this course and other DLI materials for their own communities. Course content and learning objectives follow:

Modern deep learning challenges leverage increasingly larger datasets and more complex models. As a result, significant computational power is required to train models effectively and efficiently. Learning to distribute data across multiple GPUs during deep learning model training makes possible an incredible wealth of new applications utilizing deep learning.

Learning Objectives

• Understand how data parallel deep learning training is performed using multiple GPUs
• Achieve maximum throughput when training, for the best use of multiple GPUs
• Distribute training to multiple GPUs using Pytorch Distributed Data Parallel
• Understand and utilize algorithmic considerations specific to multi-GPU training performance and accuracy

Simplot A:
Building boisestate.ai: Lessons Learned from Developing a Cost-Effective Internal AI Platform for Higher Education
Presenter- Phil Merrill

As universities race to provide generative AI access to students and faculty, the cost of commercial subscriptions at institutional scale quickly becomes unsustainable. At Boise State University, we built boisestate.ai—an open source AI platform powered by AWS Bedrock. This presentation shares the practical lessons learned from developing and operating the platform, including the consumption-based paradigm shift associated with pay-per-token pricing, seven specific cost optimization strategies (from prompt caching to semantic tool filtering), and approaches for making AI institutionally aware through MCP servers and agent skills. We'll also discuss why 2026 is shaping up to be the year of the AI agent, and how progressive disclosure and codified institutional knowledge are key to building AI that doesn't just chat — but actually gets work done for your campus.

Simplot B:
English is the Hottest New Programming Language: The AI Catalyst Model for Advanced Computing Ambassadorship
Presenter- Liza Long, Ed. D.
Academic Technology Program Manager, Idaho State Board of Education | Ph.D. Student, English, Idaho State University

As "vibe coding" and natural language interfaces become standard in advanced computing, the technical "how" is increasingly decoupled from the conceptual "what" o “why.” This shift creates a critical need for academic leaders who can bridge the gap between high-performance computing (HPC) capabilities and ethical, literate application. In Idaho, we are approaching this challenge by cultivating a culture of "ambassadorship" where faculty—particularly those from the humanities—model the critical inquiry necessary to navigate this new landscape. This presentation introduces the AI Catalyst model, co-developed with BSU Nursing professor Jason Blomquist, as a framework for decentralized AI leadership. AI Catalysts at each institution serve as bridges between technical infrastructure and pedagogical practice. I will discuss how faculty with humanities backgrounds are uniquely positioned to be these ambassadors. By applying the rigor of rhetoric, analysis, and critical thinking to "vibe coding" and AI-driven research, these catalysts model for students and peers how to be the "human in the loop." The AI Catalyst model offers a scalable blueprint for Workforce Development. It demonstrates how to move beyond top-down mandates toward a bottom-up, faculty-led movement that demystifies advanced computing. By empowering humanities-trained faculty as AI ambassadors, institutions can ensure that the next generation of researchers—regardless of their discipline—possesses the sophisticated problem-formulation skills required in a world where English has become the hottest new programming language.

Simplot C:
 Fullmoon: A Modern Web Interface for Warewulf Cluster Management
Presenter- Josh Burks

High-performance computing environments demand efficient and intuitive tools to manage increasingly complex cluster operations at scale. This presentation introduces Fullmoon: a community-developed, open-source web interface for Warewulf clusters designed to improve the administrative experience without compromising control or reliability.

Inspired by Slurm Dash, Fullmoon is built on top of the Warewulf REST API and uses Next.js to deliver a modern, responsive management dashboard. It provides advanced node sorting and filtering, streamlined node lifecycle management, input validation and sanity checks, a web-based overlay file browser, intelligent auto completion for configuration fields, and clear visualization of profile inheritance. 

The session will demonstrate Fullmoon’s capabilities, discuss architectural and security considerations for web applications in HPC environments, and highlight use cases where a web interface improves efficiency compared to traditional CLI workflows. Fullmoon provides the HPC community with an open source solution that modernizes Warewulf cluster management while preserving the robustness required for production systems.

Simplot D:
Unifying Access to Distributed Data for AI and High-Performance Computing
Presenter- Floyd Christofferson, Vice President of Product Marketing at Hammerspace

Modern HPC and AI workloads increasingly depend on data that is distributed across multiple storage systems, tiers, and locations, including on-premises clusters, institutional storage, and cloud resources. While compute performance continues to scale rapidly, data access and data movement have become primary bottlenecks, limiting utilization and complicating workflow design.

This talk examines an open, standards-based approach to unifying access to distributed data for AI and HPC workloads—without requiring proprietary clients, forklift upgrades, or disruptive data migrations. Using Hammerspace as a concrete example, the session explores how modern parallel file system standards and automated data orchestration can be used to present a single, high-performance data namespace across otherwise siloed storage systems and sites.

Attendees will learn how global namespace architectures, combined with pNFS 4.2 and policy-driven data orchestration, enable linear scaling of IOPS and throughput using existing infrastructure. The result is simplified workflow design, improved data locality, and higher sustained utilization of expensive CPU and GPU resources—particularly for AI training, inference, and data-intensive simulation workloads.

Key topics include:

• Parallel Global File Systems with pNFS 4.2 – Leveraging open standards to provide scalable, high-performance access to distributed datasets without proprietary file systems.
• Automated Data Orchestration – Using policy-driven data placement and movement to align data dynamically with compute, while maintaining continuous access.
• AI and HPC Workflow Optimization – Simplifying data access across clusters and sites to reduce staging, eliminate redundant copies, and maximize compute efficiency.
   

Simplot A:
Campus generative AI services running on HPC
Presenter- Coltran Hophan-Nichols

Follow Montana State University’s journey to implement a flexible and open-source generative AI suite, backed by our Tempest HPC system. Learn how we were able to leverage existing infrastructure and open-source platforms to provide powerful tools to all our faculty, staff, and students with no direct cost or token restrictions. Takeaways and how the local service coheres with a broader AI portfolio will be discussed. 

Simplot B:
AI Biomolecular structure prediction tools a year later
Presenter- Martin Cuma

The AI biomolecular prediction field keeps moving at a fast rate. In this talk, we'll outline what has happened at Utah in this area in the past year, including setting up a more performant server for Multiple Sequence Alignment (MSA) using mmseqs2, Boltz2 and Colabfold changes, and biomolecular design with Boltzgen.

Simplot C: 
Intel and GCC Compilers: Unleash the Power of Xeon 6 
Presenter-Xinmin Tian, Intel

In this presentation, we provide insights for Intel and GCC compiler optimizations and tunings for Xeon 6 with workload code examples. A case study on performance tuning of Torch.Inductor OpenMP code on Xeon 6 will be presented to show Intel new processor benefits. 

Simplot D:
HPC Technology in a Turbulent Market
Presenter- Chris Reidy

Purchasing equipment has become very challenging, from prices changing almost daily, to rapidly growing power consumption. We take a quick walk through the current state of the market with input from our peers and vendors.

Jordan Ballroom

Simplot A: 
Compute Anywhere with Function-as-a-Service with Globus Compute
Presenter- Vas Vasiliadis

Growing data volumes, new computing paradigms, and increasing hardware heterogeneity are driving the need to execute code on diverse distributed computing resources, many of which are outside the bounds of the researcher's institution. This need may be driven by (a) the desire to compute closer to data acquisition sources, (b) exploit specialized computing resources such as hardware accelerators, (c) provide real-time processing of data, (d) reduce energy consumption (e.g., by matching workload with hardware), and (e) scale simulations beyond the limits of a single computer.

Globus Compute addresses these needs by delivering a hybrid cloud platform implementing the Function-as-a-Service (FaaS) paradigm. Researchers first register their desired function with a cloud-hosted service, they can then request invocation of that function with arbitrary input arguments to be executed on remote cyberinfrastructure. Globus Compute manages the reliable and secure execution of the function, provisioning resources, staging function code and inputs, managing safe and secure execution (optionally using containers), monitoring execution, and asynchronously returning results to users via the cloud platform.

This tutorial will describe use cases for FaaS in science and demonstrate how Globus Compute can provide a common interface and approach for portable execution across different systems. Attendees will experiment with Globus Compute on virtual machines and learn how to deploy Globus Compute on their HPC cluster or other advanced computing system.

Simplot B:
Agentic AI for Advanced Research; Data Storage; Data Management
Presenter- Earl J. Dodd, Global HPC Business Practice Leader, Principal Solution Architect
World Wide Technology (WWT), part of the Global Solutions & Architecture (GS&A) team.

Most research environments treat storage as a procurement decision. Agentic AI flips that. Workflow and storage decide whether object, file, and parallel file systems succeed or fail, and “one big, shared filesystem” often collapses under metadata-heavy orchestration.

This session presents a workflow-first approach to infrastructure design for agentic AI and workflow-based pipelines. We characterize the I/O signatures that break classic HPC defaults, including small-file fan-out, high namespace churn, checkpoint bursts, and multi-tenant contention. We then outline a tiered architecture playbook: durable object for curated corpora, high-metadata file for orchestration surfaces, high-throughput scratch for transient staging, and policy-driven movement that preserves provenance. Throughout, we use explicit decision axes, including throughput, metadata ops, latency, and durability, so teams can justify choices to leadership and align investments to measurable bottlenecks.

Simplot C: 
Building Sovereign AI Factories: A Blueprint for State-Level Economic Growth
Presenter- HPE
AI is creating clear winners and losers across economies and regions. Sovereign AI Factories offer a powerful economic fulcrum for states seeking to attract talent, investment, and sustainable growth. By pooling resources at a state level, Sovereign AI Factories unite universities, K–12 systems, corporations, research institutes, and economic development agencies around a shared AI infrastructure that no single organization could afford on its own.

This presentation will outline the three pillars of a successful Sovereign AI Factory:

Defining a Sovereign AI Factory

• Setting goals
• Building the coalition
• Promoting the benefits

Defining the Hardware Architecture

• Scaling compute, storage, and network resources
• Addressing the need for direct liquid cooling
• Key data center considerations

Defining the User Experience

• Delivering a self-service cloud experience
• Ensuring user and resource security

• Supporting AI and HPC workloads

  Simplot D: 
  Workshop on using Generative AI in an HPC environment
  Presenter- Scott Reed
  
  Generative AI, especially large language models (LLMs) can write, explain, and refactor code from natural-language intent. Used well, it speeds up everything from Python prototyping and debugging to translating messy requirements into clean, runnable scripts. In this workshop we’ll start with the basics of how LLMs behave and why “context” and constraints matter, then apply that directly to HPC workflows: generating scripts and validating them for safety and reproducibility.

  In this interactive workshop each participant will use AI to generate bash scripts and code. Bring a laptop and a gmail account to log into google colab. We will cover LLM architecture, AI agents, skills, local vs remote LLM access, sandboxing, quantization, Retrieval Augmented Generation, and context engineering. We will explore the OpenAI API and multiple models that use this interface. We’ll focus on practical HPC workflows while enforcing safety guardrails and reproducibility.
  
  
  Atrium of the Visual Arts Building
  Come and visit the Stein Luminary! Ask Staff for directions 
  
  The Keith and Catherine Stein Luminary is an all-digital museum space, producing a 
  range of immersive, interactive and sensory experiences. Combining touch-activated 
  screens and immersive projection, we deliver cutting-edge content focused on visual 
  and performing arts and cultural exhibitions for the Boise State community.

Simplot A: 
Harnessing AI: Transforming High-Performance Computing for Next-Generation Innovation
Moderator- Shelley Knuth
Panelists- Liza Long, Phil Merrill, Casey Kennington

This panel will explore the role of artificial intelligence in high-performance computing (HPC) environments, highlighting innovative applications that enhance computational efficiency and data analysis. Attendees will gain insights into future trends and collaborative strategies for leveraging AI.

Simplot B: RMACC User Facilitation Meet-Up
Moderator- Andy Monaghan

Calling all facilitators and other user support personnel! Let’s meet up in person at RMACC to discuss our successes and challenges. Topics will include experiences in user education and outreach; the use of AI for case management and other support services; meeting the needs of customers with sensitive data, and anything else that comes up!

Simplot C: System Administrator Meetup
Moderator- Michael Ennis

Meetup with members of the RMACC SysAdmin group for an informal discussion. 

Simplot D:
LoRA, RAG, RL, Agentic AI - Making sense of the different acronyms to improve LLMs and fix hallucinations
Presenter- Michael Ramshaw 

New models from leading AI organizations come out seemingly every week, but despite the constant evolution there are always hallucinations and knowledge gaps. LoRA, RAG, RL, and Agentic AI are all popular methods to improve LLM performance and this presentation will give an overview of each method as well as discuss how "performance" can be measured in an LLM context. These methods will be compared and contrasted across several criteria such as ease of use, resources required, and the theory behind them. Open source models from the huggingface repository will be primarily used as examples since they are common in RMACC institutions that may have restricted access environments. 

Simplot A: 
UV Package Manager--Get Your Sunscreen!  
Presenter- Mohal Khandewal

Simplot B: 
Practical Guide to Performance-Conscious Python
Presenter- Robben Migacz

Python is one of the most widely used languages in scientific computing, and its adoption on HPC systems continues to grow, particularly among users with limited training in HPC and performance-oriented software development. At the same time, the ecosystem of tools for high-performance Python has expanded rapidly, making it increasingly difficult for users—and the research computing support teams advising them—to identify effective strategies to improve performance. I will discuss the landscape of high-performance Python with an emphasis on decision-making: how to choose appropriate tools and approaches based on workload characteristics and performance goals, focusing on common performance pitfalls, practical tradeoffs, and guidance for selecting technologies. The content will be most useful for Python users and research computing and data (RCD) facilitators who support Python workflows on HPC systems.

Simplot C:
Deploying and Operationalizing Intel Gaudi Systems with Kubernetes for AI Workloads
Presenter- Johnathan Lee

This session will cover ASU’s experience with the recent Intel Gaudi system donation, including the planning and buildout of new data center space to support the hardware. I will discuss the technical and operational challenges involved in bringing the systems online, lessons learned during deployment, the current status of the environment, and how we plan to integrate Gaudi into our broader research computing ecosystem to support AI and other high performance workloads.

Simplot D: 
Accelerating Scale-Out HPC and Data-Intensive Research with a Modern Parallel File System
Presenter- Joe McCormick, BeeGFS

This session examines the role of modern parallel file systems in supporting scalable, data-intensive HPC environments commonly found in academic and regional research computing centers. It discusses how BeeGFS enables high-throughput, low-latency storage architectures that effectively support a wide range of workloads, including traditional simulation, modeling, data analytics, and emerging AI-driven research.
Attendees will gain insight into BeeGFS architecture and design principles, including scalable metadata services, flexible storage tiering, and integration with high-speed networks and NVMe-based storage. The session will also present real-world user case studies from research and HPC environments, highlighting practical deployment considerations, performance characteristics, and operational lessons learned.
By focusing on production deployments and real research workflows, this talk demonstrates how BeeGFS is used today to build reliable, high-performance storage platforms that scale with growing compute and data demands in academic and research-focused HPC environments.

Simplot A: 
What's on the agenda for RMACC in 2026?
Presenter- Becky Yeager

Join us to learn more about what RMACC has accomplished over the last year
and what is on the agenda for 2026 and beyond!

Simplot B: Kit Menlove- INL’s software stack buildout

Simplot C: 
Linpack and other benchmarks on heterogeneous HPC clusters 
Presenter- Nil Mu

Benchmarking is a core part of HPC operations — whether you're validating a new cluster, justifying a procurement, or hunting down a performance regression. But running benchmarks well on heterogeneous hardware introduces challenges that the documentation doesn't always prepare you for. This presentation shares our experience running the NVIDIA HPC benchmark container on A100 GPUs and submitting results to the Top500 and Green500 lists, covering the practical details of tuning HPL parameters, navigating the submission process, and the surprises we encountered along the way. We also attempted runs across mixed A100 and H100 nodes, which raised questions about how GPU generational differences affect Linpack scaling, interconnect saturation, and whether heterogeneous submissions are even meaningful. Beyond Linpack, we'll discuss our use of the OSU Micro-Benchmarks (OMB) suite as a cluster diagnostics tool, using pairwise node communication tests to identify fabric bottlenecks, misconfigured adapters, and inconsistent latency across the cluster, and more.

Simplot D:
Scalable Patent Search and Analysis Using Large Language Models with Function Calling
Presenter- Juan Jose Garcia Mesa

This work describes a scalable system for automated patent search and analysis that integrates large language models with function calling to support data retrieval and classification. The approach combines conventional data extraction from the US Patent and Trademark Office with semantic similarity search and structured function execution to enable accurate and reproducible patent management applicable to real-world institutional data analysis challenges.

Join us at The Stueckle Sky Center 
 1200 W University Dr, Boise, ID

5:30 pm- Take a picture on the iconic blue field of Albertson's Stadium, explore the Boise State Football Stadium
6:00-7:30 pm- Join us the Stueckle Sky Center for Appetizers and Drinks, sponsored by Intel