"The Art of Insight in Science and Engineering: Mastering Complexity" by Sanjoy Mahajan.

It beautifully treats estimation and problem solving techniques, illustrated by examples from science and engineering. Instead of aiming for a complete, thorough and accurate treatment of problems, its goal is to teach shortcuts to sacrifice some accuracy for much reduced effort. This is a refreshing change to academia where rigor is often pursued at all cost. But in the real world rigor rarely matter, and simplifications are almost always worthwhile, especially initially since we can always refine models if required.

I first read it as an undergraduate and use the estimation and problem solving techniques from it almost daily. Though well hidden, the pdf is available for free on the website of the publisher.