Ask HN: How much time do you spend on deploying your ML model?

Hi. How much time do you spend on deployment of your ML models (e.g. to web as an API, to Android, iOS, etc.)?

I recently had to deploy a rather simple NN to web and Android and it looked to me like a boring, repetitive activity. For web, it was somewhat okay with copy-paste input preparation, create Flask endpoint, containerise, create instance, etc. -- I just wanted the damn model to be accessible from the outside of my computer, but it took around an hour. For Android it was a complete nightmare, since Keras transformations I've done on my input image are not ported (or I'm not aware of it).