The PhD Path: Experience and Lessons Learned (Part 2)

In my last entry, I talked a lot about the process of choosing the problems to work on and the path to and through a PhD.  In this, I close out that discussion with two topics: your dissertation and finding a job.  This "advice" is solely based on my experience as a computer science PhD student.

=== Dissertation writing ===

I'm in the process of doing this.  The most important thing here is figuring out the common thread underlying the 3 to 4 pieces of work you did over the past 3 years leading up to this point.  For almost everyone I know, this takes a very long time to do.  We have a tendency to do work on a single problem, one problem at a time, without thinking about how they all link together.  This is mainly the case because the results drive the direction of the research (if you're doing it right), so the narrative changes as your run through your experiments.
Take your time going through your results again.  The best advice I received, for writing, is to start top-down.  My advisor pushed me to get the section and sub-section titles written first, in a latex document.  It may take a few weeks to break down the set of sections and sub-section -- re-reading your papers, going through all the literature again -- reminding yourself about the fundamental contribution of your works.  Once that settles, you start to throw in text.
The text your throw in should come directly from any publication or piece of writing you've done, related to you work.  Do not massage it to fit the section headings, just throw it in.  Once you've gone through the process, it has this amazing psychological effect, where you feel like you've done a lot of work!  It's a great motivation to continue writing and also lets you see, explicitly, where the gaps are.
Work on those gaps, massage the text, and 3-6 months later, you're done!  Ok, or at least that's what i'm telling myself now that i'm in the grind :-)  But it doesn't work like this to some degree and I am planning on wrapping up very soon.
=== Assessing your job prospects ===

This is maybe the hardest part of the whole process.  What do I do with my PhD?  This can be a deeply personal question for some people.  It was for me.  I hadn't quite published significant enough work to get my pick of an academic position, I was getting tired of doing research, and I wanted to move back to a specific location -- due to a two-body problem I have (yes, a partner).
Academia.  There are many blog posts discussing going down the academic route.  Matt Welsh does a good job discussing the tradeoffs (, and is in a much better position to give advice in that area.  He also talks about the difference between a place like google and an industrial lab, like iBM Research (
For me, it personally came down to a work-life balance.  I was looking for a well-paying job, that left time for family, preferably in research, in a specific geographic area.  I tried working at a startup in the Bay Area, tried looking into development jobs, and it fundamentally came down to the criteria I layed out.  I had to try several things before really knowing.  It was a long process that start almost a year ago, when I started looking around and talking to people -- even taking a part-time position at a startup -- before i really knew what i wanted to "be when i grow up". 
For me, the process started very late.  I would recommend that you really start getting a handle on this AT LEAST 2 years before the time you think you're going to graduate.  Network.  Make contacts.  Try a few things, if that's what you need.  I rushed it, but finally settled into it.
One route that isn't often discussed is starting your own business.  I've seen this work quite successfully with folks with Berkeley PhDs (in science and engineering).  Find like-minded people, go to the business school (yes, they are needed!), and see if maybe that's a route you want to go down.  It's risky, but can pay off if you do it as diligently as your research work.
Good luck if your future endeavors!