Learning "deep" statistics, probability, and applied models of the same. My loose plan is to hack at applied stuff with Clojure and J instead of R. R is clearly the right choice for this, with SAS/Stata/SPSS right behind in terms of increasing employability.

Neither Clojure or J are particular skills for me either, so I am somewhat doing the worst approach - tackling a new project with a new set of tools . . .