Twitter has a lot of information that can be very useful if we know how to extract the relevant pieces. The main topic of the talk is to show an architecture (well tested in production). The architecture uses technologies like RabbitMQ, CouchDB, ElasticSearch, Kibana, a lot of Python and Spark Streaming with Scala. We will focus on the motivations to choose those components and how we extract the information and how we take the decisions about the obtained datasets.