RCC Spotlight: Dr. Thomas Stephan Juzek

Dr. Thomas Stephan Juzek completed an M.A. in communication science and phonetics at University of Bonn and a Ph.D. in linguistics at University of Oxford before joining Florida State University in 2022 as an assistant professor of computational linguistics. His research focuses on natural language processing, corpus linguistics and computational linguistics. 

Dr. Juzek aims to investigate the interaction between AI-generated language and human language, asking why language models favor certain lexical choices, where AI-associated language appears in real-world data and what effects that linguistic footprint may have on people and society. 

Central to Dr. Juzek’s work is building empirical evidence that can guide better model alignment and preference learning. His team studies phenomena such as lexical over-representation in LLM outputs and the emergence of AI-associated phrasing in spoken English, and they publish open datasets and tools to support reproducible research. These projects are both data- and compute-intensive, so Dr. Juzek relies on the Research Computing Center (RCC) to help with the heavy lifting for his data management workflows. In his work, he utilizes the RCC’s GPU and storage resources to fine-tune small to medium language models and coordinate API calls to cloud LLM models for inference testing and postprocessing/storage of their responses for large-scale analysis. 

Before academia, he spent several years in the tech industry working as a computational linguist and programmer—experience that informs his practical, tool-oriented approach to research. He also co-organizes the SC Artificial Intelligence and Machine Learning Seminar, mentors students in FSU’s Data Science Program and supervises graduate research assistants who use RCC resources for computational experiments. 

Dr. Juzek’s work has attracted media attention for its timely focus; outlets such as NPR and The Washington Post have featured his insights on how generative AI shapes language and the implications for model alignment.