The expansion of the World Wide Web to include information that is generated by physical devices with embedded sensing and actuation capabilities entails a surge of high-frequency real-time data that is mostly published without further processing in its raw form. To derive "smart" decisions from this data and thus use it to enable a "smart world" requires the distilling of more abstract, higher-level knowledge from it. In this paper, we propose the concept of a computational marketplace as a framework to enable the analysis and aggregation of real-time data. Here, multiple tiers of hyperlinked algorithms from different providers interact to refine data within computational graphs, which are linked structures of cascaded processing steps. We present an analysis of the key constraints on such a framework and provide a corresponding implementation as well as results from evaluations in an experimental use case scenario.