Today’s infrastructure to support Big Data science applications follows traditional control-centric approach wherein behaviour, and not data and data operations, is the primary organizing construct of its design. This limits the potential data-handling capabilities of such infrastructures. For example, it hinders explicit handling of data at various system layers (e.g., network and file system) to satisfy multi-domain requirements, such as security, performance and resource-aware data management for reducing operational costs. Central to these challenges is the fact that software tools and infrastructure to support collaboration have evolved in a piecemeal fashion. Data management technologies such as data-grids enable sophisticated operations to integrate data from multiple administrative domains into one single abstraction. Cloud models such as Infrastructure as a Service (IaaS) facilitate the rapid deployment of networked virtual infrastructure (i.e., Clouds) and fast data transfers. On one hand, these technologies do not address the Big Data challenge by working independently. On the other hand, they present complex opaque APIs and use different resource abstractions that hinder their integration, thus preventing data from playing a central role in making decisions in an automated fashion. This workshop brings together system researchers, practitioners and domain scientists with expertise and interest in Big Data Science to explore novel data-driven approaches in developing and deploying software designs and infrastructure. We will focus on capturing research that seeks to take a holistic and integrated approach to data, infrastructure and resource management; domain science applications that can benefit from these novel data-centric software infrastructures; and experiences that help us navigate the problem space.