Democratizing the Creation and Use of Biomedical Data Visualization

Speaker: Dr. Sehi L'Yi
Harvard University

Title: Democratizing the Creation and Use of Biomedical Data Visualization

Date: Monday, 24 November 2025

Time: 11:00am - 12:00noon

Venue: Lecture Theater H (Chen Kuan Cheng Forum), near lift 27/28, HKUST

Abstract:

Biomedical data, such as genomics and electronic health records, hold immense potential to advance our understanding of human health, evolution, and population diversity. Despite rapid advances in artificial intelligence (AI), fully automated analysis remains limited by the scale, complexity, and domain‑specific nuance of biomedical data. Interactive data systems, such as visual analytics, can address these limitations by keeping humans in the loop and leveraging their perceptual and cognitive strengths. However, existing systems are often too rigid, supporting only narrow tasks, or require significant programming expertise, limiting their adoption by domain experts.

My research aims to close this gap by designing interactive data systems that scale to large biomedical datasets and the diverse needs of domain experts. In this talk, I will first present Gosling, a domain-specific language that lowers the barrier to creating and using biomedical data visualizations and enables new opportunities for AI-assisted visualization authoring. Next, I will introduce Chromoscope, a scalable visual analytics system for cancer genomics that builds on Gosling to support efficient exploration of large-scale data. Finally, I will present Blace, an AI-powered visualization authoring tool that effectively combines a large language model with a Gosling-based code editor, allowing domain experts without programming expertise (e.g., clinicians, experimentalists) to easily create complex visualizations. I will conclude by discussing future directions toward end-to-end optimization of visual analytics workflows and composable, AI‑ready interactive data systems.

Biography:

Sehi L'Yi is an NIH K99 Postdoctoral Fellow in Biomedical Informatics at Harvard University. Before joining Harvard, he received a Ph.D. in Computer Science and Engineering from Seoul National University. His research bridges visualization, human–computer interaction, and biomedical informatics, with the goal of accelerating data interpretation, analysis, and communication within biomedical domains through scalable and AI-powered interactive data systems. His work has been published in leading venues across multiple disciplines, including Nature Methods, Bioinformatics, PLoS Computational Biology, VIS, TVCG, CHI, and UIST. His research has been recognized with honors, including NIH K99/R00 Pathway to Independence Award, Best Paper Honorable Mention at IEEE VIS, and Best Abstract Award at ISMB BioVis. His visualization toolkits have been widely adopted in academia and industry. For example, his visualizations are integrated into cBioPortal, one of the most commonly used cancer genomics platforms.