More than one variable

Slopes in several directions.

Partial derivatives

When a quantity depends on several inputs, you can ask how it changes with respect to each one while holding the rest still. That is a partial derivative.

The gradient collects them into a direction of steepest increase.
Going the opposite way is gradient descent, which is how models are trained.

Why it matters

Almost every real system has many inputs: price, temperature, time. Multivariable calculus is what lets you reason about them together.