Work with deep learning frameworks, model hubs, hosted APIs, local runtimes, experiment trackers, vector stores, and GPU environments. Compare managed models with open-weight deployment using cost, control, privacy, and maintenance criteria.