Research

How BIOL 172 Brings Supercomputing into Biology 

BIOL 172 at Santa Clara University combines molecular biology with Python programming and supercomputing, letting students simulate proteins and model diseases like COVID-19. Dr. Michelle McCully designed the course to give biology students practical, career-ready coding skills.

Filling the Blind Spots in Brain Imaging

Meet Sarah Anjum, a Computer Science PhD student using WAVE HPC's GPU resources to develop AI models that reconstruct full 3D MRI brain scans — enabling radiologists to detect tumors and abnormalities with greater precision than ever before.

Computational Breakthroughs in Graph Theory: Simon MacLean’s Work on Tarsi and Stanley Conjectures

Simon MacLean uses the WAVE HPC cluster to investigate major graph theory problems, including Tarsi’s Conjecture and Stanley’s Tree Isomorphism Conjecture. By developing new algorithms and leveraging high-memory computing resources, he extended computational verification of Stanley’s conjecture from 29 nodes to 33.

Cracking the Protein AI “Black Box”: Neel Mukavilli’s Research on Interpretable Protein Design

Neel Mukkavilli’s research uses high-performance computing to make protein design AI more interpretable by analyzing how models like Protein MPNN make decisions. Using the WAVE HPC cluster for AI training and molecular dynamics simulations, the project aims to turn accidental protein stability discovered by AI into intentional, controllable design for biotechnology and medicine.

Reservoir Computing Architecture as Neurotrophic Computing

This research explores reservoir computing as an efficient alternative to traditional recurrent neural networks by training only a simple readout layer. Simulations using electrical network reservoirs and Python-based tools enable scalable machine learning architectures for large datasets.

Learned Image and Video Coding

This research develops deep learning methods for image and video compression using models such as CNNs, GANs, RNNs, and transformers. Using WAVE facilities, the VIP Lab trains large-scale AI models to improve video coding efficiency for applications like streaming, surveillance, and autonomous systems.

Deep Learning Approaches for Video Prediction

Explores lightweight deep learning approaches for video prediction, using transformer-based architectures combined with CNNs and LSTMs to efficiently model spatial and temporal dynamics. The goal is to enable accurate future frame prediction on resource-constrained devices while leveraging faster training and long-range dependency modeling.

Theoretical and Computational Mechanics

Uses computational fluid mechanics and machine learning to study how microorganisms swim and to design artificial microscopic swimmers. The work supports biomedical applications such as targeted drug delivery and microsurgery.

DIANES: A DEI Audit Toolkit for News Sources

DIANES is a DEI audit toolkit that analyzes news sources to measure diversity in who is quoted and represented in journalism. Using WAVE-powered data pipelines, the project helps newsrooms evaluate and improve source diversity at low cost.

Reconstructing Two-Level Fluctuators Using Qubit Measurements

Simulates qubit signal noise caused by two-level fluctuators and reconstructs their properties using measurement data and optimization algorithms. Using WAVE’s HPC resources, the team runs large parallel simulations to help identify noise sources in quantum computing systems.

Housing and Urban Heat: Assessing Risk Disparities

This research analyzes how housing type and neighborhood factors influence heat risk in San José, finding higher risk in multifamily rentals and communities of color. Results highlight the need for targeted heat mitigation policies, especially improving access to air conditioning.

Molecular Modeling of Protein-DNA Interactions

Protein modeling uses computational simulations to study how protein structure and movement affect function, helping scientists compare natural and engineered proteins. Using WAVE facilities, researchers simulate protein-DNA interactions to predict binding strength and understand protein stability in challenging environments.

Deep Learning Approaches for Video Prediction

Explores lightweight deep learning approaches for video prediction, using transformer-based architectures combined with CNNs and LSTMs to efficiently model spatial and temporal dynamics. The goal is to enable accurate future frame prediction on resource-constrained devices while leveraging faster training and long-range dependency modeling.