Far ultraviolet (FUV) imaging of the aurora from space provides great insight into dynamic coupling of the atmosphere, ionosphere and magnetosphere on global scales. To gain quantitative understanding of these coupling processes, the global distribution of auroral energy flux is required, but the inversion of FUV emission to derive precipitating auroral particles’ energy flux is not straightforward. Furthermore, the spatial coverage of FUV imaging from Low Earth Orbit (LEO) altitudes is often insufficient to achieve global mapping of this important parameter. This study seeks to fill these gaps left by the current geospace observing system using a combination of data assimilation and machine learning techniques. Specifically, this paper presents a new data-driven modeling approach to create instantaneous, global assimilative mappings of auroral electron total energy flux from Lyman-Birge-Hopfield (LBH) emission data from the Defense Meteorological System Program (DMSP) Special Sensor Ultraviolet Spectrographic Imager (SSUSI). We take a two-step approach; the creation of assimilative maps of LBH emission using optimal interpolation, followed by the conversion to energy flux using a neural network model trained with conjunction observations of in-situ auroral particles and LBH emission from the DMSP SSJ and SUSSI instruments. The paper demonstrates the feasibility of this approach with a model prototype built with DMSP data from February 17-23 2014. This study serves as a blueprint for a future comprehensive data-driven modeling of auroral energy flux that is complementary to traditional inversion techniques to take advantage of FUV imaging from LEO platforms for global assimilative mapping of auroral energy flux.