Considering how to wrap this up as software - pipeline vs package vs pipeline creation package?

[njtierney]

Consider ways to make all of this run in a reasonable time, and think about ways to run this as software.

For example, do people need a pipeline, or do people just want a couple of functions that wrap all of this up. The key point with this is choosing a level of abstraction that makes the computation accessible and reasonable with a thought to what someone might do if they have a new dataset