UCLA

Small data are the digital traces that we produce regularly when interacting with services or devices. These data are rich, multi-modal, and often personal in nature. This thesis describes the design space of small data systems, enumerating the concerns in acquiring and deriving actionable insights from the data. An idealized architecture is presented, with spe- cific realizations of that architecture, and comparisons between these realized applications, comprising the majority of the work. While the systems discussed have a common core of primitives, the work discusses the peculiarities of each application domain and how these differences eventually lead to separate application stacks. The work closes with observations on ways to continue the exploration of this space – specifically toward small, self-contained stacks where sharing is entertained only when necessary – that draws on the work done so far and the lessons learned in that process.