What is the benefit for the customer: A user of such an app can recgnize things on his mobile phone without having to send the data to the server for processing. This might be relevant for apps that need a fast object recognition, or don't want to send to much data around.
What is the benefit for the business: The business owner might save a lot of money doing the recognition on the phone instead of on the server. Also faster response times.
How much MD did you use to build it: 8MD.
Which data is it based on: MNIST dataset (open source)
Which technologies were used: CoreML, Keras
Which things could be build with this POC if you had more time: Build a CoreML model for something where it actually makes sense to use the detection on the mobile device.