Lately, at Sovrn, we had an AI Hackathon the place we had been inspired to experiment with something associated to machine studying. The Hackathon yielded some unbelievable tasks from throughout the corporate. All the things from SQL question turbines to chatbots that may reply questions on our merchandise and different unimaginable work. I assumed this may be an excellent alternative to be taught extra about Apple’s ML instruments and possibly even construct one thing with actual enterprise worth.
A couple of of my colleagues and I teamed as much as play with CreateML and CoreML to see if we may combine some ML performance into our iOS app. We bought a mannequin educated and built-in into our app in a number of hours, which was fairly wonderful. However we rapidly realized that we had a couple of issues to unravel earlier than we may truly ship this factor.
- The mannequin was hefty. It was about 50MB. That’s lots of house to take up in our app bundle.
- We wished to replace the mannequin with out releasing a brand new app model.
- We wished to make use of the mannequin within the internet browser as properly.
We didn’t have time to unravel all of those issues. However the different day I used to be exploring the Vapor internet framework and the thought hit me, “Why not deploy CoreML fashions on the server?”
