The appearance of material defects during the manufacturing of composite structures is a relatively common occurrence. These defects can lead to premature structural collapse. The effectiveness of diagnostic techniques in detecting imperfections depends on the size of the structure and on the sensitivity of the technology. It is particularly challenging to detect small defects in very large structures, even though these frequently safety critical. It is therefore important to design and optimise these large structures with material defects in mind. Designing a structure with an optimisation framework can lead to overall better design solutions [1]. Structural optimisation methods enable the evolution of arbitrary design variables in a way that fulfils some objective and constraint functions. A particularly versatile structural optimisation method is topology optimisation, whereby the domain changes a function of some design metric (for example overall stiffness or energy release rate). In this work, we show an implementation of a topology optimisation methodology within a modelling framework designed for very large composite structures (Fig. 1). The methodology has explicit boundary-tracking capabilities [2] and relies on a formulation centred on the energy release rate of a crack or debond. We demonstrate how the optimisation methodology can handle debonds and optimise the structure in a way that minimises their effect. Finally, we apply the developed methodology to a multiscale wing-box model, studying a stringer run-out detail with a kissing bond defect. Furthermore, we show that the method can decrease the energy release of the kissing bond by changing the stringer run-out shape.