Knowledge of human’s force capacities enables the design of physical Human-Robot Interaction (pHRI) workspaces. As measuring force capacities for all postures is time consuming, predicting force capacities from a subset of measurements performed in a limited number of postures is crucial. The force capacities can be described as a convex polytope by means of a personalized musculoskeletal (MSK) model (Skuric et al. 2022). However, the tuning of a MSK model is difficult due to the high number of parameters. Thanks to its constraint-free nature on the optimization function, a genetic algorithm is implemented to find a MSK model parameter set, which fits and predicts force polytopes.