Minggang Li

I am an undergraduate at UC Berkeley studying Computer Science, Applied Mathematics, and Statistics. I work on interpretable machine learning systems for clinical speech and biological foundation models. I am currently a researcher with the Berkeley AI Research (BAIR) Speech Group and the NASA AI4LS program, and will join Visa as a Software Engineering Intern in Summer 2026 working on authentication and identity infrastructure.

Email  /  CV  /  GitHub  /  LinkedIn

Research

My research focuses on interpretable representation learning and structured semantic modeling for speech and biological data. More broadly, I am interested in how language and other high-dimensional signals can be encoded into representations that capture semantic structure while remaining robust across domains and datasets.

Working on interpretable phonological–semantic representation learning for primary progressive aphasia (PPA) subtype classification from clinical speech (WAB Picnic picture description task). Developing transformer-based modeling components and structured content-unit extraction systems for automated cognitive decline tracking.

Developing automated benchmarking pipelines for encoder-based representation learning on large-scale bulk RNA-seq datasets to study biological responses to microgravity and radiation. Evaluating cross-dataset generalization across heterogeneous transcriptomic repositories including NASA GeneLab, GEO, and ARCHS4.

Developed structured representation learning pipelines for backward-planning and semantic reasoning tasks from behavioral data. Implemented interpretable modeling experiments for decision-process prediction.

Experience

Last updated April 2026. Design adapted from Jon Barron's website.