Accomplishments for FY25-26

Research Computing

During FY25–26, Research Computing advanced the University of South Carolina’s research cyberinfrastructure through sustained operations, strategic modernization, expanded AI and GPU capabilities, secure research computing support, research software engineering, training, outreach, and interdisciplinary project collaboration.

The team supported large-scale research workloads across Theia and Hyperion, delivered millions of compute jobs and more than 118 million CPU-hours, expanded GPU-enabled research capacity, strengthened secure computing and cloud architecture for regulated data, supported REDCap growth and modernization, delivered a high-impact workshop program, and contributed to major grants, publications, presentations, and statewide cyberinfrastructure planning efforts.

Research Computing’s work supported faculty, students, staff, research centers, external collaborators, and multi-institutional initiatives across public health, neuroimaging, digital humanities, artificial intelligence, cybersecurity, bioinformatics, historical archives, mosquito surveillance, and secure health research.

Research Computing Operations and HPC Support

Research Computing continued to operate and support USC’s primary high-performance computing platforms, including Theia and Hyperion, while maintaining strong service continuity and user responsiveness.

Key operational accomplishments included:

Sustained HPC platform availability at an estimated 99.5% to 99.9%, supporting continuous research activity across multi-node computing environments.
Supported approximately 3.07 million jobs across Theia and Hyperion during the common comparison window of May 1, 2025 through May 31, 2026.
Delivered approximately 118.17 million CPU-hours to USC research workloads.
Supported nearly 230,000 GPU-hours on Theia, including significant use of A100 and H200 GPU resources.
Maintained Hyperion as a major production resource, supporting approximately 2.49 million jobs and 97.42 million CPU-hours.
Continued Theia’s growth as the university’s newer advanced computing platform, supporting approximately 594,912 jobs, 21.85 million CPU-hours, and modern CPU, GPU, AI, and machine learning workloads.
Supported approximately 352 unique job-submitting users on Hyperion, with an average of approximately 153 active users per month.
Supported H200 adoption on Theia, including approximately 6,727 jobs, 58 active users, and 23,209 GPU-hours.
Maintained strong queue responsiveness on Theia, with approximately 79.9% of jobs starting within 30 seconds and 90.4% starting within one hour.
Provided user support for account provisioning, job failures, performance optimization, software installations, storage issues, data transfer workflows, and research application troubleshooting.
Resolved 804 of 812 assigned ServiceNow tasks.
Onboarded 317 new researchers to USC research computing clusters, including 251 on Theia, 59 on Hyperion, and 7 on Bolden.

The combined HPC environment provided both continuity and modernization, with Hyperion continuing to absorb large-scale established production demand while Theia supported growth in modern CPU, GPU, AI, machine learning, and accelerated research computing.

AI, GPU, and Emerging Technology Infrastructure

Research Computing advanced USC’s AI strategy through infrastructure development, applied AI systems, GPU resource planning, AI education, and support for emerging research use cases.

Major accomplishments included:

Designed and deployed an AI Hub proof of concept, providing researchers with centralized access to large language models, AI assistants, and AI development tools.
Implemented and integrated Open WebUI, LiteLLM, vLLM, and supporting GPU infrastructure to enable scalable, multi-user AI services.
Developed AI infrastructure strategy for local and hosted models, GPU resource management, AI governance considerations, and future enterprise AI capabilities.
Supported generative AI, multi-agent systems, retrieval-augmented generation, secure AI workflows, and AI education for research and institutional use cases.
Delivered NVIDIA Deep Learning Institute workshops for students, faculty, and researchers.
Supported NVIDIA certification and developer-community engagement, including contributions to professional certification efforts in agentic AI and generative AI LLMs.
Presented and submitted applied AI work related to transcript processing, public health chatbot architecture, academic advising RAG systems, and LLM-based grant proposal evaluation.
Supported AI-focused research and development in high school transcript processing, public health resource support, institutional advising, grant review benchmarking, and controlled evaluation of multi-agent LLM systems.
Collaborated with vendor partners and the Research Computing team to plan the integration of 72 donated A100 GPUs into USC research AI and machine learning workloads, including liquid-cooling infrastructure planning.
Supported AI Day and AI Summer Camp through technical presentations, demonstrations, and outreach.

These efforts positioned Research Computing as a central partner for responsible, scalable, and research-aligned AI adoption across the university.

Secure Research Computing, Compliance, and Cloud Architecture

Research Computing strengthened the university’s secure research computing posture through cybersecurity assessment, cloud architecture, regulated data support, and compliance-focused modernization.

Key accomplishments included:

Developed a comprehensive Research Computing Information Security Assessment benchmarking SEC and R1 institutions, providing leadership with insights into program maturity, staffing, governance, and data protection practices.
Supported Data Management Plan Self-Assessment Questionnaire recertification efforts aligned with CMS data privacy safeguard requirements.
Contributed to CMMC readiness discussions focused on access control, incident response, authentication, and security awareness.
Supported secure infrastructure modernization, including high-performance computing environments, change control processes, and controlled-access research systems.
Designed and supported secure cloud-based research environments for HIPAA-regulated, CMS Medicare/Medicaid, and Prisma Health research data.
Architected hardened AWS environments for sensitive health research, including virtual desktops, on-demand GPU computing, encrypted shared storage, identity-based access control, audit logging, and HIPAA-eligible generative AI services.
Deployed foundational HIPAA workload infrastructure using air-gapped AWS services, S3 cryptographic exfiltration controls, EFS shared storage, IAM role-based access, and CloudTrail audit logging with seven-year retention.
Produced technical security documentation for Prisma Health data use applications, covering encryption, network isolation, access control, and audit capabilities.
Coordinated Transit Gateway connectivity, AD Connector integration, and Duo MFA to support centralized researcher authentication through university SSO.
Developed audit logging, monitoring, and SIEM integration strategies to improve security visibility, compliance readiness, and incident response.
Evaluated secure research computing environments for HIPAA-regulated, FERPA-protected, and other controlled-access data.

Research Computing’s secure infrastructure work directly supported the university’s ability to manage regulated research data while enabling faculty research activity.

REDCap Modernization and Research Application Support

Research Computing supported REDCap modernization in partnership with the Office of Research, with a focus on security, scalability, reliability, and long-term maintainability.

Accomplishments included:

Supported REDCap version and module upgrades to improve security and reliability.
Reviewed REDCap architecture and planned container-based approaches to improve growth, maintenance, and operational resilience.
Supported increased REDCap demand, including peak usage of up to 800 concurrent users.
Helped support growth from approximately 1,200 to 1,740 active users, a 45% annual increase.
Supported growth in first-time visitors from approximately 1,600 to 2,400, a 50% annual increase.
Continued collaboration with the Office of Research on REDCap service delivery, project management, and operational coordination.

This work helped ensure REDCap could continue serving a much larger user base than originally anticipated while improving security and long-term sustainability.

Research Software Engineering and Open Science

Research Computing expanded the university’s research software engineering capacity through open-source software development, neuroimaging infrastructure support, scientific visualization, and grant collaboration.

Major accomplishments included:

Advanced USC’s open-science neuroimaging software capabilities through sustained contributions to NiiVue and related research tools.
Supported the NIH-funded NiiVue web-based neuroimaging visualization platform, which is used by major neuroscience projects and external research communities.
Reviewed, tested, and merged community pull requests, triaged issues, and supported progress toward NiiVue’s 1.0 release.
Designed and implemented a dual-backend WebGPU and WebGL2 performance benchmark suite for NiiVue, documenting WebGPU speedups of up to approximately 19 times over WebGL2.
Extended support for large-scale and cloud-native imaging through IIIF-backed volumetric image serving, DICOM whole-slide imaging, and OME-Zarr adapters.
Contributed to NiiVue UIKit improvements, including build-system fixes, text rendering, colormaps, and themable controls.
Supported development and testing of BIDSvue, a cross-platform desktop application for BIDS-standard neuroimaging data conversion and organization.
Contributed to a U24 BRAIN Initiative proposal focused on scalable imaging and open neuroscience.
Supported peer-review activity in medical imaging research.
Submitted an NIH/NIMH R50 Research Specialist Award resubmission totaling $445,028 over three years to sustain research software engineering support for neuroimaging teams.

These activities strengthened USC’s research software portfolio and expanded the university’s visibility in national open-science and biomedical research communities.

Research Facilitation, Faculty Support, and Campus Integration

Research Computing provided direct research facilitation support to faculty, postdocs, graduate students, research groups, and campus partners.

Accomplishments included:

Met with researchers to assess computational, data, software, cloud, workflow, and storage needs.
Recommended appropriate resources, including HPC, DataHub storage, software environments, cloud services, and national cyberinfrastructure resources.
Coordinated onboarding for new research groups, including account provisioning, documentation, training, and NSF ACCESS guidance.
Advised researchers on reproducible workflows, FAIR data principles, data lifecycle planning, and research data management.
Served as a liaison between researchers and HPC operations, storage, networking, cybersecurity, and cloud teams.
Documented and escalated user issues related to performance bottlenecks, software installation, data transfer, and workflow optimization.
Supported RCD service documentation, onboarding materials, best practices, and knowledge base development.
Participated in evaluation of new hardware, software, cloud, and hybrid computing solutions.
Supported campus-wide RCD strategy initiatives related to AI and machine learning infrastructure, regulated research data, secure computing, and data-intensive science.

Research Computing also deepened campus partnerships with University Libraries, the Office of the Vice President for Research, the Graduate School, the Big Data Health Science Center, and research centers and institutes across campus.

Workshops, Training, and Instruction

Research Computing delivered a broad training and outreach program supporting practical research computing skills, AI education, HPC onboarding, data transfer, and reproducible workflows.

FY25–26 training accomplishments included:

Delivered 18 workshops, including 11 unique workshop topics.
Reached 438 registrants and 206 attendees, representing a 47% attendance rate.
Delivered 511 attendee-hours of instruction.
Increased instructional attendee-hours substantially compared with FY24–25, largely due to the introduction of NVIDIA Deep Learning Institute workshops.
Coordinated SHARPGrads sessions with University Libraries and the Graduate School.
Facilitated NSF-sponsored workshops offered by the Pittsburgh Supercomputing Center.
Scheduled and supported NVIDIA Deep Learning Institute workshops.
Delivered or supported training in HPC data transfer, Globus, Open OnDemand, Linux command line, shell scripting, MATLAB, Git, R, Python, AI, and accelerated computing.
Maintained workshop schedules, registration pages, Panopto recordings, and Research Computing web content.
Supported the conversion of Research Computing departmental policies and procedures into ServiceNow knowledge base content.
Provided individual and small-group consultations on HPC usage, SLURM, Open OnDemand, Jupyter notebooks, data transfer, workflow debugging, and software environments.

The workshop program continued to serve as a primary entry point for students, faculty, and staff learning to use USC’s research cyberinfrastructure effectively.

Outreach, Conferences, and External Engagement

Research Computing represented USC at campus, regional, and national events, strengthening the university’s profile in research computing, AI, cybersecurity, and cyberinfrastructure.

Campus outreach included participation in:

AI Day
New Faculty Orientation
Discover USC
OctoberBEST teaching fair
USC Research Centers Fair
Student Orientation Fairs
Student Job Fairs
Big Data Health Science events
SC-SHARE planning activities

External engagement included:

PEARC25 participation, including Campus Champions activities and RCD Nexus Day.
SC25 planning and booth coordination for the South Carolina Research Computing Consortium.
SC-SHARE NSF regional planning grant event support.
NAIRR Pilot Annual Meeting participation, including NAIRR Secure sessions.
Savannah River National Laboratory AI and security conferences.
Campus Research Computing Consortium participation across researcher-facing, data-facing, and strategy/policy-facing groups.
Campus Champions leadership participation and national engagement.

These activities expanded USC’s role in statewide and national research computing conversations.

Grants, Publications, Presentations, and Sponsored Research

Research Computing contributed to grants, proposals, publications, presentations, and sponsored research projects across multiple domains.

Major areas of support included:

NSF cyberinfrastructure, MRI, ACCESS, and CSSI-related proposals.
NIH data and computing components.
NAIRR and NAIRR Secure planning.
DOE and other federal agency opportunities.
Institutional infrastructure proposals.
Multi-institutional research collaborations.
ACCESS science gateway renewals supporting mosquito egg identification, SnowVision, and scRNA-seq visualization.
NSF CSSI proposal development for insect biodiversity and education research.
U24 BRAIN Initiative proposal development.
NIH/NIMH R50 research software engineering proposal development.

Research Computing-supported work contributed to publications and presentations in areas including:

Machine learning for archival film analysis.
Mosquito abundance and breeding habitat prediction.
AI-driven mosquito surveillance.
Automated high school transcript processing.
Safety-first public health chatbot architecture.
Institutional RAG systems for academic advising.
LLM-based grant proposal review benchmarking.
Historical imagery super-resolution.
Face-swapping detection using steganography.
Computer vision for mosquito egg counting.

These outputs demonstrate the team’s direct contribution to scholarly productivity and externally visible research impact.

Major Research Project Support

Research Computing provided in-depth computational support, software development, data engineering, project management, and student mentoring for several major research projects.

Public Health and Mosquito Surveillance

Research Computing supported NIH-funded mosquito surveillance and control work in collaboration with USC Public Health, Baylor College of Medicine, Florida International University, and other partners.

Work included:

Development and support of ACTT Study computing systems for mosquito analysis and control.
Support for web and mobile applications used across multiple institutions and regions.
Use of Hyperion, ACCESS resources, storage, and AI workflows for mosquito population analysis.
Applied machine learning for mosquito abundance prediction and breeding habitat analysis.
Contributions to publications and conference presentations in mosquito control, public health, and scientific computing.

Digital Humanities and Historical Film

Research Computing supported the NEH-funded AEO-Light project in collaboration with University Libraries.

Work included:

Development of an intelligent video player for historical films.
Integration of AI models for text detection, splice detection, video processing, and film analysis.
Support for a film archive web portal containing more than 2,000 films.
Use of Hyperion GPU resources, storage, ACCESS resources, library digital infrastructure, and AWS S3.
Research and development in steganography, video provenance, and historical media analysis.
Contributions to presentations at PEARC, Supercomputing, ACM Multimedia, the Library of Congress, and related venues.

Bioinformatics and Functional Genomics

Research Computing supported bioinformatics and functional genomics workflows across campus.

Work included:

Data transfer and processing support using DataHub and Theia.
Collaboration with the School of Medicine on research data workflows.
Support for the Functional Genomics Core, including data pipelines, data movement, archiving, cluster optimization, storage optimization, and performance improvements.

Data, Metrics, and Reporting

Research Computing continued to improve data collection, metrics, and reporting capabilities to better measure service impact and support strategic decision-making.

Accomplishments included:

Maintained the Research Computing Tableau site.
Maintained the Research Computing database server.
Continued development of ETL workflows integrating multiple data sources.
Aggregated and aligned data from HPC users, HPC jobs, grants, publications, faculty and staff records, student records, ORCID, workshop registration, and attendance systems.
Used Open XDMoD data from Hyperion and Theia to support workload analysis.
Began integrating publication sources such as OpenAlex and Web of Science.
Used LibCal registration and attendance data to assess training impact.
Supported metrics development for reporting, service planning, and impact assessment.

Student Employment, Mentoring, and Workforce Development

Research Computing supported student hiring, mentoring, and workforce development activities.

Accomplishments included:

Mentored research students and graduate students supported through Research Computing activities.
Represented Research Computing in Division of IT student hiring efforts.
Posted job descriptions, interviewed candidates, and supported hiring of student employees and summer interns.
Managed student employee budget and actuals processes.
Maintained domain workstations and technical resources for student employees.
Supported student worker contributions to on-premises and off-premises websites.
Participated in student hiring committee meetings and related workforce coordination.

These efforts contributed to both operational support and the development of the next generation of research computing professionals.

Strategic Impact

In FY25–26, Research Computing continued to serve as a critical institutional capability for USC’s research enterprise. The team delivered reliable high-performance computing operations, expanded GPU and AI infrastructure, strengthened secure research computing, supported regulated data environments, modernized research applications, enabled interdisciplinary research projects, contributed to grants and publications, and delivered practical training for the university community.

Research Computing’s work directly advanced USC’s ability to support computationally intensive research, data-intensive science, AI-enabled discovery, secure health research, open-source software development, and statewide cyberinfrastructure collaboration.

Division of Information Technology

Research Computing