William Catt

I build the whole thing — taking a vague question through to a model, then to something you can actually click on. I care about evaluating that work properly, which includes reporting the experiments that didn't pan out. Most of it lives in NLP and LLM systems, predictive modelling, and simulation.