Abstract This work presents a systematic approach toward the design of Organic Rankine Cycles (ORC) for the generation of power from multiple heat sources available at different temperature levels. The design problem is approached in a mixed-integer non-linear programming (MINLP) formulation where an inclusive and flexible ORC model is automatically evolved by a deterministic algorithm for global optimization. The basic building block of the model is the ORC cascade which consists of a heat extraction, a power generation, a condensation and a liquid pressurization section. The aim of the optimization is to determine the optimum number of ORC cascades, the structure of the heat exchanger network shared among different cascades, the operating conditions and the working fluid used in each cascade in order to identify an overall ORC structure that maximizes the power output. The approach is illustrated through a case study which indicates that a system of two waste heat sources is best exploited through two interconnected ORC utilizing different working fluids.