The alternative - which I did 2 weeks ago at work - is to use Hive through AWS EMR. Then it's just SQL all the way instead of reinventing the wheel like this article recommends.