Uncertainty Quantification
Generative language models are not always accurate, and they struggle to reliably gauge their intrinsic uncertainty, motivating ancillary systems that do so on their behalf. My work often views this through the lens of distribution property estimation.
Model Behavior and Interpretability
Generative language models can behave in unexpected ways, for reasons that are notoriously opaque. Responsible deployment of machine learning systems warrants careful examination of such behavior and potential failure modes.
Data Science Applications: Health Sciences
It has been suggested that ~30% of all data generated is attributable to the healthcare sector. Public health and biomedical informatics are rich areas of application for natural language processing and network science.
Data Science Applications: Logistics and Engineering
Through my experience at the Logistics Management Institute (LMI), I have had the opportunity to collaborate with industry practioners on practical problems involving technical language processing and simulation.