In 2018, Idaho National Laboratory built DeepLynx, a data warehouse designed to organize large amounts of engineering and scientific data. As INL's projects grew more complex and AI became more central to their work, the original platform couldn't scale to meet new requirements. DeepLynx Nexus was built to address these limitations. Rather than replicating and storing large datasets, Nexus catalogs metadata and relationships about data that lives in existing systems. Think of it as a smart catalog that doesn't just tell you what data exists and where to find it, but also explains how different pieces relate to each other, where they came from, and what they mean. This rich context is exactly what AI agents need to actually do useful work. This presentation provides a hands-on walkthrough of getting Nexus running locally and cataloging your first datasets. We'll cover installation, configuration, creating a data schema, and a brief overview of Apache Airflow, a common ETL adapter architecture we use to bring metadata into Nexus. By the end, you'll have a practical understanding of how Nexus works and how it's being used to support lab initiatives.
Advanced computing environments are rapidly adopting AI to accelerate, automate analysis, and streamline decision-making. That speed is useful, but it introduces a problem IT teams and research leaders often underestimate: synthetic confidence. As AI-generated outputs begin to look authoritative, organizations risk accepting flawed results, weak explanations, or manipulated content with less scrutiny than they would apply to a human analyst. This session explores how trust is formed, misplaced, and exploited in AI-enabled computing environments, particularly where high-performance computing, sensitive research, and collaborative workflows intersect. Attendees ill walk away with a practical framework for improving verification culture, reducing automation bias, and building security controls that protect not just systems, but judgment itself.
The NSF Office of Advanced Cyberinfrastructure (OAC) supports over 25,000 researchers and students through an integrated ecosystem of computing, data, networking, and software infrastructure. This presentation provides an overview of OAC's major initiatives and resources relevant to the research computing community, including the NSF Leadership-Class Computing Facility (LCCF) — a distributed, national-scale system entering production in FY2027 with 2,000 GPU nodes, 390PB of flash storage, and an 800PB archive — and the National Research Platform (NRP), which aggregates GPU, CPU, and storage resources across 84 organizations for research and education. The presentation also covers the ACCESS program for allocating advanced computing and data resources, and the National AI Research Resource (NAIRR) Pilot, which connects researchers and educators to AI computing infrastructure, datasets, and training. Highlighted NAIRR projects span battlefield medicine, agricultural resilience, Alzheimer's disease prediction, and deepfake detection. Finally, the presentation surveys upcoming funding solicitations including IDSS, CICI, CSSI, FAIROS, Future CoRe, and TechAccess: AI-Ready America, offering pathways for institutions to engage with and contribute to the national cyberinfrastructure ecosystem.
Most faculty are still early in their AI journey. Yet the future is already arriving: small, self-learning AI models that power autonomous agents capable of continuous improvement, collaboration, and generating entirely new capabilities for teaching and research. This talk paints the compelling vision of where we are headed: a shift from consuming AI to building with it, from fragile trillion-line codebases to intelligent, living systems that replace much of traditional software while enabling personalized tutoring, interactive simulations, adaptive research assistants, and new forms of discovery. We explore why small, efficient models (10 to 30 times cheaper, on-device fast, privacy-first) represent the practical foundation for campus-scale agents, how self-learning agents will transform classrooms and labs, and what petri-dish environments for safe experimentation could look like. A live demonstration of projectEureka shows the foundation where this future can be built, seamlessly running custom AI models and autonomous agents across on-prem and cloud environments. Join us to be inspired and motivated to prepare your institution for the new era of building AI models and autonomous agents in teaching and research.
Cloud computing has become a foundational enabler for academic institutions seeking to scale research workloads, modernize curricula, and reduce infrastructure overhead. This session explores how colleges and universities are leveraging Amazon Web Services (AWS) to address key challenges in higher education — from burst-capable HPC clusters for computational research, to cost-effective storage for growing datasets, to AI/ML platforms that bring cutting-edge tools into the classroom. We will examine practical patterns for deploying research computing environments on AWS, including integration with schedulers like Slurm via AWS Parallel Computing Service, and strategies for managing multi-account environments across departments and research groups. We will also highlight the AWS Open Data program, which provides free access to large-scale public datasets — enabling researchers and students to focus on analysis rather than data acquisition and hosting costs.
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.
Please join Mark III and NVIDIA at RMACC for an overview, walk-through, and live demo of building an AI agent app quickly and easily using NVIDIA NIMs and Blueprints. The first segment of this session will focus on a build of a simple AI agent bot built with a Llama 3 NIM. The second segment will focus on a live build and demo of a more advanced AI agent build with an NVIDIA Blueprint (powered by multiple NIMs). Lastly, the session will show how to quickly and easily finetune a model before being served up for API consumption by apps via NIMs. This session will be of interest to participants of all levels and skillsets looking to deploy AI services into their existing and new cloud-native and modern apps and research as quickly and effectively as possible.
The release of Snakemake 8 and 9 introduced breaking changes that forced HPC-dependent bioinformatics workflows to fundamentally rethink their cluster integration strategy — most notably, the removal of the long-standing `--cluster` flag in favor of a modern executor plugin interface. This talk walks through the real-world migration of DETECT, a simulation-based de novo mutation detection pipeline, from Snakemake 7 to Snakemake 9 on a SLURM HPC cluster. We cover the technical challenges encountered along the way: replacing inline cluster submission strings with the `snakemake-executor-plugin-slurm`, restructuring resource declarations into profile-based `set-resources` blocks, implementing automatic partition selection, and adding submission rate limiting to prevent SLURM socket timeouts at scale. Beyond the executor change, the migration prompted a broader modernization — containerizing GATK with Apptainer to eliminate conda instability, resolving Python environment path issues in shell rules, and building operational tooling for log analysis and interactive setup. Attendees will leave with a concrete migration roadmap and reusable patterns applicable to migrating from any Snakemake 7 workflow to Snakemake 8 and 9running on SLURM.
