Question on putting all things together.

I am pretty new with ML, but I have done few small projects on ML using scikit-learn, opencv, etc. I've done online classes like Andrew Ng's ML as well. I can say I am quite familiar with how things work or the processes in general.

But I still do not understand how to put things together especially on the deployment of ML. Currently, let's say I'm doing a predictive engine or some sort, after done with training the samples, now I have the model of the prediction and I can start pumping in new sample to test.

In real project, how do people actually persist the information of the model after it is done with training? I've done googling, some suggest to use pickle (im using python btw). What are some other methods to actually store the information without the need to run the training sets again. This is the thing I am really confused. Most of examples only show a simple script that run training set and then run the prediction.

Thanks!