Hm... I had an idea about this once. It'd start as a research study on professors in universities and analyze their teaching philosophies, the trends of the schools where they got their credentials, their internalized expectations of students, and their cognitive preferences, as a means to group teachers by certain clusters (not MBTI). From there incoming students would be matched w/ professors who shared similar cognitive styles and objective philosophies. But it never left the random brainstorming stages for obvious reasons, like overflow, workplace readiness, time constraints of the study, and the possibility of limited selection.
I think when it comes to primary and secondary education, the point of most schools is to create good employees; not good scholars. By college, it depends on the university, and the discipline of study, and even then, the schools still have an "eye on the door". The problem with teaching to a certain style would be the fact that the person would have inherent weaknesses when it came time to relate to different people who'd been taught by different styles, and also the person would be limited by whatever blindspots the teaching style afforded (I think this maps back to that Socratic teaching conundrum).
If your goal is to teach the same skillset of standardized materials using different cognitive styles or preferances, such as visual-spatial, logical-mathematical, auditory, etc. then you'd have a better launching pad than with personality/MBTI. But then AGAIN, you'd run into trouble attempting to convey certain subject matter through a constricted medium.