After learning the basic syntax, reading some non-trivial code is a fast way to learn a language. We can also learn how to design a library/software during reading others' code.
I have following lists.
A Chess program in OCaml by Tomek Czajka.
Hal Daumé has written several machine learning libraries in Ocaml. Including decision trees, logistic regression and SVM. All of them are near-production-quality code.
A Chess Game Analysis program in F# in Microsoft Research.
The above three are my favorites. Will you suggest some other sources? General purpose open source software are good, specialized open source like the three I list here are even more welcome.