Christan Grant

The UF Data Studio develops AI theory and systems for data that current models cannot process. The research program addresses three questions. First, what can distributional analysis reveal about model behavior that aggregate metrics cannot? We develop mathematical frameworks that decompose prediction errors into their full distributional structure, yielding diagnostics invisible to standard evaluation. Second, how do we extend NLP and vision beyond phonetic text and standard imagery? We build methods for encoded visual communication systems, including semasiographic manuscripts, woven codes, and symbolic art traditions that encode information outside written language. Third, how do intelligent agents and humans collaborate in complex data environments? We study multi-agent LLM dynamics, human-AI annotation systems, and anomaly detection in urban sensor networks.