Many financial risk and valuation applications are – at their heart – dataflow computations. Few developers, however, have tools that enable them to easily express the dataflow relationships inherently present in these applications. I developed a dataflow-centric grid computation system. This system leveraged annotated F# to allow developers to create “live” dataflow applications where outputs were recomputed on-demand whenever an input to the graph changed. Perturbations of the dataflow graph in this system allowed efficient expression and computation of risk metrics. Moreover, this system contained a table-centric, bi-temporal data store that enabled consistent treatment of time-series data, corrections to data, and risk perturbations.