It's a lifecycle

The model is the smallest part.

A real ML project is a loop: frame the problem, get data, train, evaluate, deploy, monitor — then repeat. Training the model is one step among many, and rarely the hardest.