Overview
The Catalytic Project-Based Grant program is an interdisciplinary initiative designed to accelerate high-risk, high-reward research by providing faculty with the early support needed to explore bold, computationally intensive ideas and pilot exploratory projects. By empowering faculty across all departments to generate proof-of-concept data and establish compelling preliminary results using the High Performance Computing (HPC) cluster, the program effectively positions researchers to secure subsequent external funding from major agencies like the NSF or DOE. Ultimately, this program expands campus-wide expertise in computational and visualization scholarship, fostering interdisciplinary collaboration and enhancing the research capacity of faculty and students who rely on SCU’s cutting-edge computational resources.
Academic Year 2025–2026 Awarded Grants
Exploring Opportunities for Performance Optimization in Quantum Circuit Simulations
Faculty: Younghyun Cho, Assistant Professor, Computer Science and Engineering
Project Overview: This research investigates algorithmic and system-level optimization opportunities within quantum circuit simulations on classical HPC systems. By profiling leading simulation frameworks, the project aims to improve the scalability and efficiency of full-state and approximated simulations. Experiments utilizing SCU’s WAVE HPC and AMD GPU clusters will explore hybrid simulation methods that are directly applicable to the emerging domains of quantum machine learning and circuit synthesis.
Towards a New Generation of Hybrid Density Functionals Using HPC
Faculty: Robin Grotjahn, Assistant Professor, Chemistry and Biochemistry
Project Overview: Dr. Grotjahn’s project seeks to develop a new class of hybrid Density Functional Approximations (DFAs) to achieve higher accuracy in modeling molecular interactions without prohibitive computational costs. The research utilizes the Turbomole quantum chemistry suite on the WAVE HPC for large-scale parameter optimization. If successful, this work could enable a transformative generation of hybrid DFAs, significantly improving the precision of electronic excitation predictions in the field of quantum chemistry.
Mentorship for All: Multi-Agent Multilingual Long-Form Video Question Answering for Mentorship Applications
Faculty: Oana Ignat, Assistant Professor, Computer Science and Engineering
Project Overview: This project develops a novel multi-agent AI framework designed to extract pedagogically meaningful question-answer pairs from long-form mentorship videos. The system processes over 150 hours of video data across multiple languages, including English, Romanian, Hindi, and Marathi. By building a unique multilingual QA dataset and evaluating it against advanced baselines, the research aims to democratize access to mentorship through accessible, AI-driven educational technologies.
Computational Modeling of Porous Materials for Biofuels
Faculty: Maryam Mobed-Miremadi, Associate Professor, Bioengineering
Project Overview: This initiative leverages HPC resources to accelerate multiphase fluid-dynamics simulations critical to the design of biomedical devices and drug-delivery systems. Utilizing computational fluid dynamics (CFD) and finite element modeling, the project predicts microdroplet behavior and optimizes bioreactor performance. The research involves developing simulation workflows to analyze complex transport phenomena at microscale resolutions, providing the foundational data needed for next-generation bioprocess systems.
Federated Reinforcement Learning for Client-Centric Multi-Link Optimization in WiFi 7
Faculty: Krishna Kattiyan Ramamoorthy, Assistant Professor, Computer Science and Engineering
Project Overview: Dr. Ramamoorthy’s research focuses on the simulation and optimization of 802.11be (WiFi 7) protocols, with a specific emphasis on Multi-Link Operation (MLO). Using the WAVE HPC for large-scale network modeling, the research seeks to improve data throughput and significantly reduce latency in the next generation of wireless communications. The complexity of these simulations, which model high-density traffic across multiple frequency bands, requires the sophisticated parallel processing capabilities provided by the WAVE architecture.
Decoding Visual Engagement: Computational Discovery of Image-Driven User Engagement Mediators
Faculty: Shunyao Yan, Assistant Professor, Marketing; Zijing Zhang, Assistant Professor, Marketing
Project Overview: This interdisciplinary project investigates how visual stimuli influence consumer attention and decision-making using deep neural networks trained on large-scale video datasets. By leveraging WAVE HPC resources, the team extracts multimodal features—including eye movement, facial emotion, and scene context—to predict engagement metrics. The goal is to build interpretable AI models that reveal visual attention dynamics, providing high-resolution insights for marketing, media analytics, and neuroscience.
Academic Year 2026–2027 Awarded Grants
Exact Universal Sparse Attention via Quantization-Aware L2-Norm Pruning
Faculty: David C. Anastasiu, Associate Professor, Computer Science & Engineering
Project Overview: This project addresses the computational bottlenecks of long-context large language models by developing an exact universal sparse attention mechanism based on l2-norm pruning. Utilizing WAVE HPC GH200 nodes to implement specialized CUDA kernels, the team aims to achieve bit-exact quantization that maintains accuracy while accelerating real-time inference for applications like video anomaly anticipation.
Non-Adiabatic Molecular Dynamics on WAVE HPC
Faculty: Robin Grotjahn, Assistant Professor, Chemistry and Biochemistry
Project Overview: This proposal focuses on implementing local hybrid functionals for non-adiabatic molecular dynamics to accurately simulate the localization of electrons in photoexcited mixed-valence compounds. By running high-performance simulations on WAVE HPC using the Turbomole package, the project aims to establish a new methodological standard for studying light-harvesting and molecular electronics.
Benchmarking the Clinical Reliability of Healthcare LLMs for MH/SUD Care
Faculty: Haibing Lu, Professor; Jingbo Hou, Assistant Professor; Information Systems and Analytics
Project Overview: This project benchmarks the reliability of large language models in behavioral health by constructing a clinically faithful synthetic dataset grounded in real-world electronic health records. Leveraging HPC for large-scale inference, the team evaluates multiple LLMs on complex tasks like diagnosis assessment and medication adjustment to identify safety gaps and improve clinical decision support for mental health care.
Design of Thermostable Cellulases
Faculty: Michelle E. McCully, Associate Professor, Biology
Project Overview: This research employs LLM-based protein design and molecular dynamics simulations on WAVE HPC to develop thermostable cellulases for efficient biofuel production. By applying state-of-the-art packages like NOMELT and ProteinMPNN, the team aims to design enzymes capable of withstanding industrial high-heat conditions while maintaining functional integrity during biomass conversion.
Deep Reinforcement Learning for Opinion Maximization in Social Networks
Faculty: Venkatesh Srinivasan, Professor; Smita Ghosh, Mathematics and Computer Science
Project Overview: This project adapts deep reinforcement learning to recommend social network links that maximize favorable opinions within leader-follower models. Using WAVE HPC for repeated simulations and GPU-based model training, the team will develop a framework to improve opinion equilibria in both static and dynamic temporal networks.