Background

We use Jenkins at BrandVerity as our Continuous Integration server. When someone pushed a change that would fail the automated tests Jenkins would send an email to the dev team about it. In practice it was easy for people not to notice these emails and the tests could remain broken for a big part of the day (which blocked the deployment of code to production). So I wanted to create a mechanism for the team to be able to quickly see if the tests were passing or not.

BrandVerity has been doing dev hackdays every 6 weeks for a long time now. For this particular hackday, I was interested in playing with unsupervised anomaly detection, a concept I first encountered in this really great Andrew Ng machine learning class on coursera.