Research infrastructure, interdisciplinary collaborations, and independent technical experiments.
Projects are continuing bodies of work rather than individual publications. Each page brings together the associated papers, software, articles, talks, and other outputs.
Open-source infrastructure for reproducible out-of-distribution detection research.
Applying data mining, visualization, and language models to questions and collections in social-work research.
An end-to-end experiment in preparing data and training a 358-million-parameter German GPT-style model from scratch.
Using data mining, language models, and visualization to examine German news coverage and parliamentary speech.