In this paper, we propose a mixed-integer linear programming (MILP) model for building a resilient, cross-trained workforce in mixed-model assembly lines, incorporating learning effects from both training and regular assembly operations. The model minimises the estimated overload generated by workers through their assignment to either training or assembly tasks across multiple shifts. Worker processing times are reduced as a function of the tasks they perform or train on, following an assembly learning curve. We present a modelling framework for integrating learning curves into a MILP formulation using binary variables that track the number of repetitions each worker performs on each task. This allows the learning curve to be linearised piecewise and directly incorporated into the model. An illustrative example is provided to evaluate the model’s ability to develop a workforce composed of both novice and experienced workers. Results show a cross-training pattern, achieved by rotating workers between workstations and training, leading to a reduction in estimated overload.