Gradient Descent finds the minima of cost function, by using the derivative of the cost function w.r.t parameters. Without applying Gradient Descent, we cannot train any model in Machine Learning.