Solution space methods facilitate robust design and concurrent engineering by defining decoupled intervals for design variables that ensure system-level feasibility. However, conventional techniques, typically relying on single hyperrectangles or convex polytopes, implicitly assume a connected, approximately convex feasible region. This assumption breaks down in nonlinear engineering problems, such as crashworthiness or vibration control, where the performance landscape is inherently multi-modal. This paper introduces the Multi-Modal Solution Space (MMSS) framework to address this limitation. The approach integrates optimization-guided exploration with adaptive density-based clustering to partition the design space into distinct feasibility islands. Subsequently, it computes local, box-shaped solution spaces via volume maximization. Boxes may have partially overlapping design variable intervals while remaining geometrically distinct; complete overlap triggers redundancy removal. Application to design problems demonstrate that MMSS identifies distinct design modes compared to single-box method.