RMACC HPC Symposium 2025

This presentation explores AWS latest developments in AI infrastructure and tools designed specifically for research communities. We'll examine the transformation of SageMaker Studio, the introduction of the Build on Trainium program, and the strategic positioning of Amazon Bedrock alongside SageMaker AI. This session will highlight how these integrated solutions address the evolving needs of researchers, from individual projects to large-scale collaborative efforts. Live demonstrations of SageMaker Studio and Bedrock API integration will showcase the practical application of these tools in research environments.

AI assisted protein interaction modeling, pioneered by AlphaFold and RosettaFold, has become more diverse both with respect to the programs that do it, and how users run these programs. In this talk, we will cover the programs that are supported at the University of Utah, namely Alphafold2, Alphafold3, Colabfold, Boltz1, RFDiffusion and other tools from the Baker lab, the choices we have made with their deployment, and our experiences with using them. With respect to the ways to run, we will go over the standard SLURM scripts to run Alphafold in two stages (CPU only MSA search, GPU accelerated inference), use Colabfold server for faster MSA search, and using Google Colab running on compute nodes for interactive modeling in a notebook interface. Attendees should leave this talk with ideas how to set up and support these tools and contacts to UofU staff for further questions.

A universal quantum computer can be used as a simulator capable of predicting properties of diverse quantum systems. Electronic structure problems in chemistry offer practical use cases around the hundred-qubit mark. This appears promising since current quantum processors have reached these sizes. However, mapping these use cases onto quantum computers yields deep circuits, and for pre-fault-tolerant quantum processors, the large number of measurements to estimate molecular energies leads to prohibitive runtimes. As a result, realistic chemistry is out of reach of current quantum computers in isolation. A natural question is whether classical distributed computation can relieve quantum processors from parsing all but a core, intrinsically quantum component of a chemistry workflow. Here, we incorporate quantum computations of chemistry in a quantum-centric supercomputing architecture, using up to 6400 nodes of the supercomputer Fugaku to assist a quantum computer with a Heron superconducting processor. We simulate the N2 triple bond breaking in a correlation-consistent cc-pVDZ basis set, and the active-space electronic structure of [2Fe-2S] and [4Fe-4S] clusters, using 58, 45 and 77 qubits respectively, with quantum circuits of up to 10570 (3590 2-qubit) quantum gates. We obtain our results using a class of quantum circuits that approximates molecular eigenstates, and a hybrid estimator. The estimator processes quantum samples, produces upper bounds to the ground-state energy and wavefunctions supported on a polynomial number of states. This guarantees an unconditional quality metric for quantum advantage, certifiable by classical computers at polynomial cost. For current error rates, our results show that classical distributed computing coupled to quantum computers can produce good approximate solutions for practical problems beyond sizes amenable to exact diagonalization.

Publication of research datasets is now a requirement of most funding agencies and journals. Data curation is the process of ensuring that these datasets are findable, accessible, and usable. In the era of Big Data, the generation of datasets with sizes on the order of 100s of gigabytes and larger is increasingly common. Such large datasets create challenges for both the curation and publishing of data as they often cannot be accessed on standard computer hardware or hosted in traditional online repositories. This presentation provides an overview of a collaborative process between the CU Boulder Libraries and CU Boulder Research Computing in which high-performance computing infrastructure is used to curate and publish gigabyte- and terabyte-scale datasets in a manner that makes them accessible to the research community.

The Front Range GigaPop (FRGP) is a longstanding network in Colorado that connects universities, government agencies, and schools to Internet2, peering exchanges, and the broader Internet through shared fiber-optic, routed and packet-switched infrastructure.

A recent NSF grant is funding the expansion of FRGP to the Western Slope, providing 10Gbps connections to Mesa State Metropolitan State University in Grand Junction (MSU), Western Colorado University in Gunnison (WSU), and the Rocky Mountain Biological Laboratory in Gothic (RMBL).

This presentation will highlight FRGP’s services and showcase its role in advancing cyberinfrastructure (CI) across the region.

The University of North Dakota (UND) Genomics Core has launched GenomEX 2.0, the first comprehensive and user-friendly bioinformatics platform powered by Oracle Cloud Infrastructure. This innovative platform enables biologists to seamlessly install over 13,000 bioinformatics tools, generate and execute custom code or command lines, and receive real-time guidance from an AI-based bioinformatics assistant—all through intuitive, one-click processes.
To support the computing requirements for any bioinformatics tools, the platform is powered by the Oracle Cloud Infrastructure that provides fully secured (built-in security features and compliance certifications), personalized (adjustable CPU/GPU numbers & memory/storage capacity), dedicated (resources available 24/7 without any queue) and customizable (users have administrator rights) cloud-based high-performance computing environments at unbeatable pricing.
Through the combined expertise of the UND Genomics Core and Oracle, GenomEX 2.0 emerges as a powerful and unique bioinformatics platform, providing every biologist with the freedom to explore biological data independently, regardless of their coding proficiency.

How can your research teams leverage and apply high performance computing to supercharge your own efforts for research momentum to get to outcomes faster?

When generative AI (GenAI) was first introduced, our AI team identified the proposal process as a top use case for GenAI augmentation. Our AI experts set out to build a GenAI-powered application to help the Proposal team streamline the complex RFP process. The goal was to build an intelligent assistant that accelerated RFP qualification, helped proposal managers quickly grasp the key aspects of lengthy RFP document sets, and searched multiple databases for relevant information to automatically generate a quality response draft.

The now successful internally developed GenAI solution significantly shortens the Proposal team's response time, allowing them to answer to more proposals and spend more time on response quality and win strategies. This use case will be expanded to support the whole process of winning customer engagement, from drafting Statements of Work (SOWs) to proposal PowerPoint presentations.

In the spirit of exploring the art of the possible, we will share a live demo of capability in use for creating proposals and responses for project funding.

This presentation explores AWS comprehensive solutions for the research community, addressing three critical challenges: meeting evolving compliance requirements like the new NIH NIST 800-171 standards, delivering scalable high-performance computing, and simplifying data management. We'll examine how the AWS Secure Research Environment (SRE) addresses complex security needs, compare managed AWS ParallelCluster Service with self-managed options, and showcase tools like the Globus S3 Connector for streamlined data handling. Join us to discover how AWS empowers researchers to focus on innovation while maintaining security, performance, and efficiency at scale.

This talk shares the story of an NSF-funded experiential learning opportunity for undergraduate and graduate students at RMACC institutions. Students developed practical skills in HPC system administration by learning from and shadowing CU Boulder Research Computing staff. A total of 17 students participated across two in-person experiences and took part in various aspects of system design, deployment, and teardown.

Come and learn more about Quantinuum and their compute offerings

Perhaps management is in your future. Or you have recently been cast in that role and you would appreciate some suggestions. Attend this session to collect some useful references, and discuss the importance of our role and maximizing the impact of our teams

In the presentation, we would share info on topics such as metrics, our user survey, and some other approaches. As part of the talk, we would like to engender a discussion and exchange of info about what other sites do to measure the effectiveness of their HPC environments.

Managing data at scale in high-performance computing (HPC) environments requires efficient storage and retrieval strategies. Automated tiered storage solutions enable seamless migration of aged data to lower-cost archival tiers while maintaining accessibility. Enriched metadata—spanning tagging, search, discovery, and data provenance—enhances data usability and long-term value. This approach not only optimizes storage costs but also empowers researchers with better data discovery and reuse. Real-world HPC use cases demonstrate how metadata-driven workflows streamline research, ensuring that critical datasets remain accessible and actionable over time.

AI workloads present new challenges for traditional HPC architectures, particularly as compute demands outpace I/O performance, power constraints tighten, and budgets remain stretched. The need for high-throughput, low-latency data access at extreme scale is forcing a re-evaluation of storage architectures to maximize efficiency without incurring unsustainable costs.

This session will explore how standard Linux and open storage technologies enable AI and HPC workloads to achieve parallel file system performance on commodity hardware—without requiring specialized infrastructure. Topics include:
• Scaling AI Storage with Standard Linux: Leveraging NFSv4.2 advancements, including pNFS and FlexFiles, to enable parallel I/O at extreme scales.
• Next-Generation SSD Integration: How embedding parallel file system capabilities into SSDs reduces data movement overhead while maximizing power efficiency.
• Accelerating AI with Localized Storage on GPU/CPU Servers: Techniques to optimize checkpointing and ephemeral data storage, reducing I/O bottlenecks and overall infrastructure costs.

The session will feature real-world examples of how these innovations are being deployed today to drive high-performance AI workloads while addressing power and cost constraints.

This presentation highlights the rapid growth in Field Effect Transistor (FET) technology, with modern GPUs containing over 200 billion transistors each. As transistor counts soar, so does power consumption—Nvidia’s shipped GPUs alone consumed over 14 TWh in 2023. With global data center energy demands projected to triple, the presentation advocates for transitioning data from power-hungry disk storage to more energy-efficient systems. This shift offers significant power savings, lower carbon emissions, and cost reductions—providing a scalable path to support future computing growth without overloading existing power infrastructure.

This workshop will act as a continuation of last year's presentation and discussion on quantum-centric high-performance computing (QCHPC). We will again look at realistic instances in the current noisy intermediate-scale quantum (NISQ) -era, how these (and other use cases) might scale with quantum hardware, and strategies for integration of quantum resources with HPC. We will consider approaches to putting these hybrid paradigms into practice; Symposium members are encouraged to contribute efforts made by their HPC departments, both from emulation/simulation and hardware integration perspectives. The workshop is intended as a space to explore ideas, share experiences and gather knowledge to advance quantum-centric HPC in our region.

Fuzzball, from CIQ, is a new way to access and manage HPC resources, whether they're on-prem, in the cloud, or both. Fuzzball is API-driven, container-forward, and offers co-equal web and command-line interfaces. With its well-defined workflow definitions, including first-class support for container management and data ingress and egress, Fuzzball workflows are portable to both local clusters and cloud resources.

In this session we will give an overview of Fuzzball, describe some of its potential use cases, and highlight some of its unique features, including a visual workflow builder, portable workflows, and its templated workflow catalog.

As research demands grow and infrastructure ages, some institutions are turning to the public cloud to supplement or replace traditional on-prem systems. This talk shares the journey of launching a public cloud pilot for research computing at a small state university. We’ll explore the drivers behind the shift—including scalability, agility, and cost transparency—and walk through key decisions, from selecting a cloud provider to identifying test users. Drawing from real-world experience six months post-launch, we’ll cover what’s worked, what’s surprised us, and what we’re planning next. Attendees will leave with a practical roadmap and lessons learned to guide their own cloud adoption efforts—whether starting small or scaling up.

CUDA-Q is an open-source quantum development platform orchestrating the hardware and software needed to run useful, large-scale quantum computing applications. The platform’s hybrid programming model allows computation on GPU, CPU, and QPU resources in tandem from within a single quantum program. CUDA-Q is “qubit-agnostic”—seamlessly integrating with all QPUs and qubit modalities and offering GPU-accelerated simulations when adequate quantum hardware is not available.

CUDA-Q extends simulation tools far beyond the NISQ-era—charting a course to large-scale, error-corrected quantum supercomputing.

Field-Programmable Gate Arrays (FPGAs) are gaining traction in research for their ability to deliver high-performance, energy-efficient computing across a range of domains—from machine learning and data analytics to signal processing and scientific simulations. However, integrating FPGA workflows into a shared university compute cluster presents unique challenges in terms of hardware management, toolchain support and user access. This session will explore the practical aspects of supporting FPGA applications in a multi-user academic environment. We will cover available FPGA platforms, commonly used development workflows (such as Xilinx Vivado and Vitis, Intel Quartus and OpenCL, and HLS), and the architectural and administrative considerations for cluster integration. Real-world use cases will illustrate how researchers and academics are leveraging FPGAs, and we’ll share lessons learned in enabling productive FPGA development. Attendees will gain insight into both the technical setup and the support models that foster a thriving FPGA user community on campus.

The next frontier in AI advancement isn’t just about algorithms—it’s about unlocking the wealth of hidden insights trapped within millions of files in HPC environments. While organizations focus on model architectures, the true bottleneck often lies in discovering and preparing relevant data buried in vast storage systems.
This presentation, featuring MetadataHub and a live demonstration, will reveal how intelligent metadata extraction and management transforms unstructured data into AI-ready assets by:

• Uncovering Hidden Context: Live metadata extraction demonstrating how MetadataHubcaptures content and contextual value, revealing unexpected connections between research datasets and enabling new AI training opportunities that would otherwise remain hidden.
• Automating Data Discovery: Demonstrating how MetadataHub automates metadata tagging to identify valuable training data across petabyte-scale storage, reducing data preparation time by up to 90%.
• Enhancing Model Quality: Exploring how rich metadata captured by MetadataHub improves AI model performance by providing better context and enabling more relevant training data selection.
• Scaling Efficiently: Showcasing metadata-driven automation with MetadataHub that optimizes data pipeline efficiency and resource utilization, including GPU/CPU performance, across HPC environments.

The session will highlight a real-world success story from the Zuse Institute Berlin, where MetadataHub unlocked 200 PB of previously underutilized research data for cutting-edge Generative AI applications. A 15-minute live demonstration will guide attendees through their journey—from data discovery to AI-ready datasets—highlighting practical challenges and solutions.
Attendees will leave with actionable strategies for implementing metadata-driven approaches in their own HPC workflows. By showcasing MetadataHub’s ability to extract content and contextual value, this session will demonstrate how metadata transforms unstructured data into a strategic advantage, accelerating AI initiatives and driving HPC innovation.