I'm currently evaluating/prototyping with it for a startup idea I have, in the high assurance/white-hat IT security space. Current favourites are Poly/ML for a compiler. For libraries, I'm interested in contract.sig & .sml from mltonlib for design-by-contract, and use-lib from mltonlib for a build system. Will probably use the event-driven net I/O library from the Poly/ML contribs page too.
I registered for the Coursera course by Dan Grossman, but Real Life has intruded so I had to drop out. A pity, it looked good. Anyway, I know a few related programming languages (e.g. Haskell, Scheme, ISLisp) so probably quicker just to read books. For books, Mads Tofte's "Tips for Computer Scientists" is a great 20-page summary if you have the background to understand it. And for more detail or if you need a gentler on-ramp, Robert Harper's "Programming in Standard ML" looks good too.
I had beel playing around with ISLisp but wanted something standardised with type-checking and support for programming-in-the-large. Specifically, I wanted some way to do object-capabilities in a mainstream programming language, I've managed to port Oleg Kiselyov's OCaml code from his "Lightweight Static Capabilities" to SML without much trouble. Also wanted design-by-contract (I looked at dependent typing instead of this, but it seemed to be more trouble than it was worth). So here we are.