Methods comprising loading a data model in-memory, providing a first graphical object of loaded data based on the data model, wherein the first graphical object represents a plurality of data sets, executing a first procedure in an inference engine based on a user selection in the plurality of data sets to generate a data subset, executing a second procedure in a calculation engine to generate a first multidimensional data cube based on the data subset to generate a second graphical object, providing at least a portion of the data subset to an external engine to perform a third procedure, receiving a result of the third procedure from the external engine, repeating the first procedure and the second procedure based on the data subset and the result of the third procedure to generate a second multidimensional data cube and to generate a third graphical object, and providing the third graphical object.