Data analytics, Spark, Hadoop and AI have become fundamental tools to drive digital transformation. A critical challenge is moving from isolated experiments to an organizational or enterprise production infrastructure. In this talk, we break apart the modern data analytics workflow to focus on the data challenges across different phases of the analytics and AI life cycle. By presenting a unified approach to data storage for AI and Analytics, organizations can reduce costs, modernize their data strategy and build a sustainable enterprise data lake. By anticipating how Hadoop, Spark, Tensorflow, Caffe and traditional analytics like SAS, HPC can share data, IT departments and data science practitioners can not only co-exist, but speed time to insight. We will present the tangible benefits of a Reference Architecture using real-world installations that span proprietary and open-source frameworks. Using intelligent software-defined shared storage, users are able to eliminate silos, reduce multiple data copies, and improve time to insight.