Rapleaf's company blog, where engineers of the data and people-profile startup wrote about Hadoop, Cascading, distributed key-value stores and hiring.
HN links it for its engineering posts, such as the EC2 versus colocation comparison, and for discussion of its unusual hiring practices and FizzBuzz screening.
I'm not sure if this is totally over the board here, but Rapleaf has a number of... unusual hiring practices. So I want to weigh in before others decide to copy their practice without thinking.
For example, the job description for an 'Amazing Software Engineer' is a little over the top.
http://blog.rapleaf.com/jobs/#engineer
- Amazing coder who takes no prisoners. Master of all things Internet. One of the best coders in existence.
- Intensely driven and proactive person.
- Extremely hard working. This is a start-up - team members work long hours.
- Quick learner and real doer. Err on execution over strategy.
- Thrive on working with A-players. Too good to spend long hours with B-players.
- Likeable person who garners respect on and off the job.
- Thrive on chaos, risk, and uncertainty.
- Should be easy to get along with, nice, fun, smart, ethical, and low-maintenance.
- Strong desire to build a more ethical society.
- Desire to be an early employee and want to be a real owner in Rapleaf’s future.
- Want to work with extremely large datasets and build portable APIs that thousands of other companies can build applications on top of.
Note: this job is really hard. You’ll be working with some of the top search engineers in the world and they are going to expect that you kick ass. We’re doing things that no one has ever done before and solving problems that have been open for years.
---------------------------------------
I think this attitude would be totally fine, were it…
The only part that's theoretical is the extent we're being tracked online but there's plenty to suggest it's "as much as possible".
- Google's targets ads based on your browsing history - that could include Search, Analytics, AdSense, DoubleClick, and a whole ton of other data they have and collect. [1]
Rapleaf tracks so much that it's actually "a challenge" not to identify people. [2]
QuantCast settled for $2.4 million for making sure their precious tracking cookies were recreated if you delete them. [3]
Facebook just tried to give apps your fricking phone number and address. [4]
Google and their referral leaking is not really "the bad guy" in all of this - there's no bad guy, just a bunch of companies consuming and analyzing as much data as they can get their hands on.
It's not evil, but is it necessary?
[1] http://googleblog.blogspot.com/2009/03/making-ads-more-interesting.html
[2] http://blog.rapleaf.com/dev/2010/07/20/anonymouse/
[3] http://www.wired.com/epicenter/2010/12/zombie-cookie-settlement/
[4] http://technolog.msnbc.msn.com/_news/2011/01/18/5868697-facebook-wont-share-phone-numbers-addresses-for-now
Rapleaf has posted a response here:
http://blog.rapleaf.com/referrer-urls-and-privacy-risks/
Though it doesn't seem to explain the details of what happened. It would have been useful for them to include the specifics of what occurred and exactly what data was erroneously transmitted to whom (a la Foursquare's post-mortem from a week or two ago.) This could be as valuable a lesson for system architects as the 4sq one was.
Yes, but now 1 sysadmin manages between 10 and 100 times more machines than they used to[1].
What is more, each machine supports more people than it used to.
[1] http://blog.rapleaf.com/dev/2008/12/10/rent-or-own-amazon-ec2-vs-colocation-comparison-for-hadoop-clusters/
[2] http://www.datacenterknowledge.com/archives/2009/12/30/how-many-servers-can-one-admin-manage/
I first heard about FizzBuzz from Hackruiter.com and then I got an email from RapLeaf (http://blog.rapleaf.com/dev/2011/09/06/code-golfing-at-the-office)
It's a very interesting concept to filter out the bad programmers using something simple like FizzBuzz. What do you guys like to give out as programming or interview questions? How effective are they?
Python: http://www.michael-noll.com/wiki/Writing_An_Hadoop_MapReduce_Program_In_Python
Ruby: http://blog.rapleaf.com/dev/?p=16
It's pretty hokey, either running through the shell or calling a REST service through a proxy, but it's been done if you're allergic to non-scripting languages.