Jeremy Howard and Rachel Thomas founded fast.ai with a vision to make artificial intelligence more accessible to the masses. In addition to developing an open-source Python library, the company offers MOOCs which enable students to develop state-of-the-art machine learning models. Much of this is accomplished with fast.ai’s results-oriented, top-down teaching method.

Like most others, formal education trained me to learn from first principles. I didn’t question this bottom-up approach too seriously until I started my first job out of school. I began my career in the flight simulation industry, and was expected to start contributing in short order. The code-base was enormous, the hardware was complex, and the hardware-software integration initially seemed like black magic. My manager guided me along a top-down approach: I was thrown into projects, but constantly encouraged to ask questions. I remember feeling bewildered most days, and often resorting to a trial-and-error approach to accomplish my work. But within a few months, I tackled enough problems to start gaining an intuitive understanding of many subsystems. Over the subsequent two years, I deepened my technical aptitude into a number of areas at a personally unprecedented pace.

My experience with the top-down approach has led me to incorporate it into many facets of my life. It allows me to learn a new subject with the bigger picture in mind at all times, and keeps me excited throughout the process. This method may create an initial hump of frustration and uncertainty, but I firmly believe that it is the most efficient process to learn a subject. I’m glad I stumbled onto fast.ai, and plan to stick with it.