This idea is great, a bit hard to do due to the equipment, but still.

A lot of times I've stumbled on the fact that we don't have flat, even surfaces. This hinders a lot of things, those microscopíc imperfections.

So I set my mind to try to find a way to get flat surfaces. 

I'll explain this idea in simple terms, but you should be trying to modify it to make it more professional. There is a moving table, and on its top a rectangular piece of paper. A mechanical pencil is placed vertically so that it draws on the paper. Carbon graphite, easy to transfer to paper.

The table moves in 2 dimensions, with very good precision. This is like a plotter, a common arduino experiment. Alternatively you may choose to move the pencil and keep the paper still, its should work too.

The thing is, you will also need to have a very precise way of measuring the carbon layer thickness on the paper. Do you know those devices people use to get sizes on a dimensional measuring machine, with a ball on the tip? I thought about three of them, forming a three-fingered hand, with the pencil on the middle of the triangle. This hand hold still while the table moves the paper. 

You would set an algorithm: move the pencil to the direction where the graphite layer is thinnest. You would measure thickness in those three directions, and give a step to the direction where there is a depression on surface, so that the pencil can fill it. This should have a control loop.

Keep doing that for a long time. With some luck, the pen will slowly get the surface even after it has completed a long trajectory on the paper.

How flat this surface will get will depend on the whole set: the graphite tracing, the precision of the table, the distance sensors. Vibrations will be an issue too if you want to get really flat. If you wanted to top this, you could use interferometry.