Google announced TensorFlow Serving today. The basic notion is simple – take your trained TensorFlow model and make it into a web service running on some scalable hardware. Predictions on demand.
The fact that this is part of the TensorFlow way of doing things is not the important bit. (In fact, truth be told, I’m finding TensorFlow to be more of a pain as we use it and I’m looking at alternatives). But the notion of deploying models behind a service API is the big idea. People should get comfy with that notion because I expect we will end up using it a lot.