Mostly we prefer to use cloud platforms for our products/applications, because of security, availability, maintenance, scalability, and cost-effectiveness. Also they provides monitoring tools to know about application performance. These tools help us to understand CPU load, memory, disk, latency, throughput, and many more of a single instance or cluster.

Very important concept for training models with private data without uploading the user’s private data into the cloud.

Data powers all of our machine learning models. TensorFlow Data Services makes managing data much easier. As developers, one of the things that we’ve discovered when we are building models is that we often have to write far more lines of code to get our data and to slice our data and to manage our data to be able to feed it for training, then we write code for the actual training itself.

We’ve seen a lot of deep building algorithms run maybe on our system and on the Cloud, but there’s something magical to getting these algorithms. Maybe a model that we’ve trained to run in your hand, on our smartphone or on a lightweight embedded processor like, an Arduino Raspberry Pi. TensorFlow Lite, an exciting technology that allows us to put our models directly and literally into people’s hands.

One of the existing things about JS is that it allows us to do neural network training and inference right there in the web browser. So, it’s really cool that a user can upload a picture to a web browser or grab an image from a webcam and then have a neural network do training or inference right in the web browser without needing to send that image up to the cloud to be processed by a server.

Maybe we have trained a model in Jupyter Notebook or in our system, but how do we take that model and have it be running ²⁴⁄₇, have it serve actual user queries, and create value? This course will teach us how to do all that using TensorFlow.