Connecting the two ideas is not straight forward, because you are dealing with different objects. OP seems to perceive this as a problem of dimensions but that will not be fruitful. Let's take a circle of radius r=1. One way to express this is y_1(x)=\pm \sqrt{1-x^2}
another way starts with a bucket:
The height of the function (respectively the z-coordinate) is the square of the distance of a point given by x,y from the origin. So z(x,y)=r^2=x^2+y^2. This is like your k(x,y,z) but we just use k(x,y).
If you set z(x,y) equal to a constant, then you cut the bucket at a certain height, and get a circle. Actually if we set it equal to 1 we get the circle given by y_1(x):
z(x,y_2)=1
x^2 + y_2^2=1
y_2=\pm \sqrt{1-x^2} = y_1
The derivative y_1'(x) gives you the growth rate of y_1 when proceeding along the x axis, which also happens to be the slope of the tangent line.
The gradient of z(x,y) points outwards in the x,y plane in the direction where the bucket grows the fastest.
So it has two components. The reason why it is normal to the tangent line on the circle which we get when we cut the bucket at a certain height is as follows:
All the points of the cut are at the same height. So the growth along this line is 0. This means that this direction is perpendicular to the direction of strongest growth or strongest decline. In a way it is the midpoint between the two. (The exact reasons lie in the linearity of tangents but I won't elaborate on this.)
To summarize: The gradient will give the normal to a constant surface of a function that lives one dimension higher. And the derivative gives the growth rate of a function with respect to a certain variable. It doesn't point anywhere, and to make it a tangent you have to discuss the equation of the tangent line.
Sorry this might be a bit dense, if you have not seen theses expressions for circles before.