Question B
I would have liked to write in Scala for this project. It appears much cleaner and stays out of the way, unlike Java's rather verbose type system. The functional programming aspects of Scala leave little code to be written, and much less to worry about with types. Furthermore, I continually hear about how Scala is being used in enterprises, and it might be a good skill to learn. In addition, being familiar with a functional programming language like Scala (or Haskell) would be beneficial for thinking about other programming languages.
Question C
I see that MLlib does not support training neural networks, just simple support vector machines (SVM's) and linear models. It would be very powerful if Spark could orchestrate the training of a neural network in a server farm armed with hundreds of GPU's; training would be insanely fast. Many data scientists would be attracted to this feature.
