For both galaxy formation and early Universe work, it's extremely important to be able to take advantage of computers to make detailed and precise calculations.
The typical approach of scientists to computing is to get the job done via the quickest method available. For simple problems, this is a pragmatic approach. But as our codes and the demands we place upon them become increasingly complex, it becomes unacceptably difficult to show that the computers are actually doing the right thing.
I'm interested in taking a much higher-level approach to scientific computing, in particular the use of ultra-high-level languages like python. As well as giving graduate-level lectures on these topics, the most visible aspect of my work in this area is the new simulation analysis library pynbody (taking over from my older package siman). Pynbody represents a collaborative effort for which I'm leading development.