Building a recommendation engine with AWS Data Pipeline, Elastic MapReduce and Spark