Computer science · Research · Education
Joe Dwyer
I study how training choices affect the efficiency of language models, and how computer science education can keep pace with a changing field.
PhD in Computer Science · Associate Dean of Software Development, ECPI University
Research focus
My doctoral research examined how changing the number of training tokens affects parameter efficiency. I use controlled experiments and compute-aware measures to ask not only how a model performs, but what the training run costs.
That work now informs my interest in practical, resource-conscious machine learning and in making software development education relevant to the tools students will encounter.
Current work
Associate Dean of Software Development
ECPI University · August 2026–present
I lead software development programs, including the design and implementation of a bachelor’s degree in AI-augmented software development. I previously served as the university’s Computer Information Science Program Director.