Here's a web app engineer perspective: delivery to the customer's laptop is likely over HTTP so the mechanism is taking customer text input, calling API's and delivering results.In its simplest form (again from a web app engineer's viewpoint) LLMs ingest text, organize that data (a very complicated process I don't understand fully, nor need to), and provide API's to output text given some set of inputs so this resembles very closely working on Elastic or other search engine technology. A caveat ...