Ask HN: What would Parse for Deep Learning look like?

I saw Ilya Sukhar's tweet about this (https://twitter.com/ilyasu/status/725357333111398405) and was wondering what that service would look like.

A few services are mentioned in the replies (e.g. https://wit.ai, https://scaledinference.com, http://algorithmia.com) but Ilya notes that he "don't see any apis". Additionally, I don't think that any of the mentioned services are as simple and easy to get started with as Parse was for someone who has little domain knowledge in ML.

For a quick prototype I focused on iOS and made a simple Cocoapod library for "plug and play machine learning" that calls my server requesting item recommendations based on the user's "like" history. The backend implements a very basic KNN.

Any thoughts are welcome on how to approach this and if anyone is interested in trying out the prototype please send me an email at zkirill ... gmail.com or reach out on Twitter https://twitter.com/zkirill

Thank you kindly for your comments!

EDIT: Fix spelling.