Liliana Royer 2024

Liliana is from ‘Ewa Beach, O’ahu and graduated from Daegu High School in South Korea. She is a senior pursuing a bachelor’s degree in Computer Science at University of Hawai’i at Manoa. Liliana believes that computer science is the backbone of modern academia and enjoys the diversity that the field has to offer. As a lover of the natural sciences, she hopes to find a career that intersects her passion for both science and technology. She seeks this intersection in the fields of software engineering, data science, machine learning, or low level programming. Liliana is a senior event coordinator for the Association of Computing Machinery (ACM) chapter at UH Manoa and a STEM Scholar in the Native Hawaiian Science and Engineering Mentorship Program (NHSEMP). Liliana currently spends her time on the weekends as an ICSparks mentor teaching students coding and web development. She also loves waterfall hikes, supporting local farmers’ markets, and immersing herself in nature as much as possible. Liliana is adamant about social justice, preserving culture, and caring for the environment.

Home Island: O’ahu

High School: Daegu High School

Institution when accepted: University of Hawaii at Manoa

Project Site: Maui High Performance Computing Center: MHPCC, Kihei, Maui

Mentor: James McSweeney

Project Title: Developing a Visualization Tool for AI Generated Satellite Images

Project Abstract: Famously, most machine learning models are “black boxes”. Even if a model has a strong performance it is hard to accurately describe how these decisions are made. This explainability is a huge topic of conversation in understanding AI ethics and how AI works. This project will focus on creating a visualization of an ML model that is used to simulate and determine the orientation (pose) of satellites. Possible visualizations include heat maps on images, computer workload graphs, or node use diagrams. This project will also assess how the number and types of images used in these ML models affect the predictive ability of the model (e.g. under or overtraining can lead to biases).