A bi-level stochastic framework is being developed which strategically coordinates open pit mining operations with downstream metallurgical operations, taking into account geological uncertainty. The computations involve two levels of optimization: (1) the outer level applies a Variable Neighbourhood Descent (VND) to determine in which planning period each mining block is to be excavated, and (2) the inner level applies the Network Simplex Method (NSM) to deduce the corresponding inventory/processing decisions, and thus to estimate the expected NPV of each excavation plan. The resulting algorithm may be an important contribution to the field of geometallurgy, as it links geostatistical data to metallurgical production. The VND and NSM have complementary features, and are therefore particularly well-suited to strategic geometallurgical planning. The VND computation considers a combinatorially large number of block orderings, resulting in numerous applications of the inner computation. However, this outer level makes only incremental transitions from one potential ordering to another, resulting in only gradual changes in the NSM input data. Indeed, the inner optimization benefits from this gradual behaviour, as it uses the results from one iteration to accelerate the following iteration. The current implementation makes considerable use of the standard C++ container templates, in conjunction with the ILOG-CPLEX libraries. This paper describes the design and function of the data structures, and presents sample computations.