Stephanie Murray

About Me

Hi, I’m Stephanie Murray, a researcher and PhD student in Computer Science at the University of Hawai‘i at Mānoa. My work focuses on machine learning, data analysis, and algorithm development, with research interests in large-scale biological datasets and phylogenetic methods.

I have experience developing analytical pipelines for complex datasets, including work in biomedical signal processing during my master’s thesis. I’m interested in creating computational approaches that make challenging, high-dimensional data more accessible and interpretable.

I enjoy working on meaningful problems, learning from new ideas, and exploring creative solutions in computer science that can have a positive impact.

Updated: October 2026

News

Thumbnail of the SHARD-DB SC26 research poster

November 2026 — SC26 poster presentation

My research poster, SHARD-DB: Scalable Hash-Based Retrieval for High-Throughput Reference Database Construction, has been accepted for presentation at Supercomputing 2026 (SC26) in Chicago. I will present this new computational biology and high-performance computing work during the research poster session on November 17. This project is distinct from the quantum machine learning research I presented at SC25.

November 2025 — SC25 poster presentation

I presented my master’s thesis poster, Advancing EEG Signal Analysis with Quantum Machine Learning, at Supercomputing 2025 (SC25) in St. Louis. The poster summarized results from my written thesis comparing variational quantum circuits to a tuned Random Forest baseline.

Full poster image (PNG)

December 2025 - Michael Bruno Award for Excellence in Microbiome Science

Received the Michael Bruno Award for Excellence in Microbiome Science through the C-MĀIKI #MahiMicrobe2025 program for a proposed Rapid ‘Ōhi‘a Death microbiome project focused on comparing microbial communities from surviving and infected trees using amplicon sequencing, computational biology, and machine learning, in collaboration with Dr. Lenore Pipes.

Rapid ‘Ōhi‘a Death microbiome project preview

Research Interests

Computational Biology: Developing algorithms and analytical methods for large-scale biological datasets, with a focus on phylogenetics and environmental DNA (eDNA).
Machine Learning: Designing and evaluating models for classification, prediction, and pattern discovery in complex, high-dimensional data.
High-Performance Computing (HPC): Leveraging parallel computing and GPU acceleration to improve efficiency in data-intensive workflows.
Data Analysis: Exploring techniques to address small, sparse, or imbalanced datasets through feature engineering and statistical modeling.
Biomedical & Life Sciences Applications: Applying computational methods to datasets from neuroscience, genomics, and environmental monitoring.

Projects

SHARD-DB: Scalable Hash-Based Approximate Retrieval for High-Throughput Reference Database Construction - Ongoing in 2026

Under the supervision of Dr. Lenore Pipes at UH Mānoa, I designed and developed SHARD-DB, a bioinformatics and high-performance computing system for constructing marker-specific reference databases from very large nucleotide collections.

Rather than using repeated alignment-first BLAST expansion, SHARD-DB indexes reference sequences with densified MinHash sketches and locality-sensitive hashing (LSH) to retrieve broad candidate neighborhoods, then performs targeted locus localization to identify and refine marker regions for downstream database construction. The implementation is designed to reduce repeated data movement through compact posting lists, query-directed metadata access, memory mapping, shared reference-region processing, and bounded-memory execution.

Current results show that performance is strongly marker-dependent: on a full NCBI NT MiFish workload, SHARD-DB used substantially less peak memory than CRUX-style merged BLAST while retrieving approximately 12× more unique candidate accessions, providing a richer pool of sequences for reference-database construction. Historical matched-subset experiments with MarVer3 16S and PR2 18S also showed large runtime advantages for SHARD-DB.

Additional SHARD-only candidates have been independently evaluated using semi-global Needleman-Wunsch alignment, and ongoing work is focused on characterizing retrieval sensitivity, precision, and parameter choices across marker families. SHARD-DB remains active, unpublished research, with a journal manuscript in preparation.

SHARD-DB genomic analysis with candidate loci

UW Master’s Thesis (Completed June 25): "Exploring Quantum Machine Learning-Enhanced Models for EEG Data Classification"

This 103-page thesis investigates the application of Variational Quantum Circuits (VQCs) for classifying EEG signals. The work integrates quantum computing methods with classical preprocessing to improve classification performance for movement-related brain activity, using both simulation and real quantum hardware tests.

Across 40 experimental runs, the VQC achieved higher precision and recall than a tuned Random Forest baseline, demonstrating that compact quantum models can match or exceed classical pipelines for certain biomedical signal processing tasks.

Methods Used: Variational Quantum Circuits, classical preprocessing (band-pass filtering, PCA, CSP), and gradient-free optimization (COBYLA) for training quantum models.

Full thesis available via ProQuest now!.

SightShift: VR Assistive Technology (Completed)

Worked in a team in conjunction with MXTReality to create a VR application simulating LHON (Leber Hereditary Optic Neuropathy) disorder using Unity and Oculus SDK. Focused on accessibility and assistive technology, promoting empathy and understanding of visual impairments through immersive experiences.

Methods Used: Unity game engine, VR development (Oculus SDK), and accessibility design principles.