Predictive Discovery Limited discovered and is developing the Bankan Gold Project, a Palaeoproterozoic shear controlled Birimian deposit in Guinea. A Definitive Feasibility Study based on the current resource model has produced a Probable Reserve of 51.6 Mt @ 1.78 g/t to be mined from three open pits and an underground mine, at an all-in sustaining cost of US$1057 per oz. In the main North-east Bankan (NEB) open pit, the mineralisation is a complex network of mineralised anastomosing shears with intervening low-grade and barren rock; this has been demonstrated on scales from macro to micro. Structural measurements of oriented core provided a coherent set of shear orientations, which enabled the interpretation of the first order and some of the second order structures as inputs to the resource model. The current mine plan is for a low-selectivity, conventional drill, blast, truck and shovel approach as interpreting the smaller scale shear network from wide spaced resource data carries significant interpretation risk. A deterministic interpretation of a trial grade control program was used to derisk the mineral resource estimate and provide realistic ore loss and dilution factors for the resource to reserve conversion; however this single interpretation is based on subjective interpretation choices, similar to the resource model and does not consider interpretation risk. Novel machine learning based approaches have been applied to the grade control data to provide a set of alternative interpretations of the shear network. This ensemble modelling framework enabled calibration and quantification of both geological interpretation uncertainty and grade control estimation risk. The outcome is a methodology that supports the deployment of a risk-based grade control strategy in routine production, as multiple scenarios can be considered in daily ore waste decisions.