The proliferation of Internet-based platforms such as online social networks and other web services, as well as the use of different user devices, has drastically expanded the plurality of data produced. The sheer volume and velocity with which data, social or not, are produced have increased multifold and the functions executed on such data have become more complex and computationally expensive. To address these new challenges, we need a new paradigm for designing distributed systems that handle such big, fast, multi-purpose, multi-device social data. This paradigm should take into account key characteristics of the data and their users-owners, and embed these properties in the design of the supporting system.

In this talk I will cover various explorations of this paradigm, by extracting user behaviors and characteristics from social data and embedding this knowledge in the design of systems responsible for managing and processing said data, or informing the design of system features for improved performance and user experience. In so doing, I will cover efforts on different types of distributed platforms such as social data management in a P2P network, content abuse management in a question-answering community, network analysis of streaming graphs, and user cost management in real-time-bidding ad ecosystems.