Most of the hard part of a model is upstream of the model. We work with teams who have run out of what public datasets can give them and need something collected, checked, or built to their own specification.
That covers gathering data that doesn't exist in a usable form yet, cleaning and labelling what a team already has, and putting together evaluation sets for measuring whether a system actually works. Scope and format depend on the project, so we'd rather talk about yours than list packages here.
Nearly everything we do is under contract, and we don't write about client work publicly. If you think there's a fit, get in touch and we'll tell you plainly whether there is.