Recurrent neural networks have recently achieved spectacular results in many different natural language processing tasks, but their utility in more practical applications is largely undocumented. For companies that follow a subscription business model, sequential data can often be found in abundance, seemingly making recurrent neural networks a perfect fit for many prediction tasks. Unfortunately, resources describing how to leverage the power of recurrent neural networks in non-language settings are generally lacking. At Red Hat, we’re using recurrent neural networks to tackle a number of different business goals, including predicting customer churn and prioritizing support cases. During this talk, you’ll learn about Red Hat’s full deep learning pipeline, from data collection (e.g., mining website logs on Hadoop clusters and pulling data from SQL databases), to data preprocessing (e.g., dimensionality reduction in scikit-learn), to prediction. By the end, you’ll have the foundation necessary to begin implementing your own recurrent neural network solutions.