A friend of a friend of a friend of mine got in contact with me and offered me the job, which was to write code for something, of which he couldn't say until I got my security clearance and accepted the job. They just wanted someone who was a good problem solver. Anyway, it turns out that the EFs have some kind weird effect on computers that causes them to behave non-deterministically, sort of like quantum computing, and I, along with about half of Google's former staff, have been working to try to find ways to either work around the problem or to make use of the effect to get better results.
In particular I've been working on the problem of DNA sequence matching. The details aren't very interesting, but one of my results was devising an O(sqrt(N)+sqrt(M)) solution to a problem that is O(N+M) under the RAM model. We're still working on figuring out a formal model for EF computing; I just kind of seat-of-pantsed it.
Unfortunately, I apparently overexposed myself to the EF, so now I'm stuck with three or four of them circling me all the time. This doesn't go over well in bars.