What is gradient descent and how does it work in the context of optimizing a machine learning model?

What is gradient descent and how does it work in the context of optimizing a machine learning model?