Also how are they different from directed edge a YC funded start up.
If they are doing predictive analysis/machine learning stuff, It makes sense to use Java, Python and C in that order since Java is what Hadoop is written in, while python and C allow very low dev time and latter provides extremely fast execution.
Using an unproven second class language like clojure shows that the team behind this project has a little experience in statistical modeling and related fields.
No one case which lanugage you use if its FP or ruby or even worse fortran. In statistical modeling it finally boils down to fast execution of some sort of huge matrix multiplication.
Really the only concern w.r.t language is readability/ compatibility with existing systems and low dev. time