One will ask, why do we want to do that? Is not deep learning and TensorFlow something that you do on massive clusters of GPU machines in big data centres? The answer to this question is obviously yes, but it's also an amazing way of offering unique user experiences that you have never been able to offer it to people before and that is by running interactively on mobile devices.

Mobile applications have lots of potential because people use it on daily basis. You can offer dozens of apps the depend on deep learning from real time translation, to predicting next words on keyboards, to help people scan their old photos with photo scan, to all the amazing stuff that you know SnapChat is doing in real time. You can also develop medical mobile applications that can help diagnosing diseases and more.

At the moment not all devices are supported by TensorFlow but we are going to only cover the Android part of it. TensorFlow at the moment supports:

• Android
• IOS
• Raspberry PI

But with the growing community of TensorFlow, you can expect the more is coming.