I would suggest starting off with something simple. The language and platform are not important.
Try to think about which topics you find interesting, ever wanted some program that does something for you? build it, even if it exists. once it is built try to think about what is the next step and how to improve it.
By going through a full development cycle: requirement->design->implementation->test->deploy
you will teach yourself how to be effective in writing code and possibly find what topics you feel a stronger connection to.
Some concrete suggestions:
Build a Mashup - use an API of some popular site you use or like and build your own application on top of that API. Some suggestions:
See the ProgrammableWeb for many many more APIs and mashup ideas.
Build an Android (mobile) app - build something simple in java that allows you to do one function you think is cool, it doesn't matter if someone else already built it, try building it yourself. see this is as an example
Build a simple desktop application - build something that does some sort of data processing, perhaps it will read data feeds from some online source and store them in a database. later it can use that information to display something interesting.
For example: when I was at the university I had the same question you have, I decided to build an application that predicts the outcome of sports games.
I first stated by scraping sports gambling sites to get the betting ratio, later on I added the ability to read league / tournament score tables and store the history of the teams. Once I had that I started adding features to the prediction engine, streak length, home strength, rank difference, venue type... I built a roadmap for my application and tried to visualize how it will look next month and next year.
Eventually I stopped working on my own project when I started working, but I can certainly say that my job interviews got progressively easier (and more interesting to the interviewer) the further along I was in my project.