Chassis strain suitable for producing multiple compounds is a central concept in synthetic biology. Design of a chassis using computational, first-principle, models is particularly attractive due to the predictability and control it offers, including against phenotype reversal due to adaptive mutations. Yet, the theory of model-based chassis design has not been put to experimental test. Here, we report two Saccharomyces cerevisiae chassis strains for dicarboxylic acid production based on genome-scale metabolic modelling. The chassis strain, harboring gene knockouts in serine biosynthesis and in pentose-phosphate pathway, is geared for higher flux towards three target products - succinate, fumarate and malate - but does not appreciably secrete any. Introducing modular product-specific mutations resulted in improved secretion of the corresponding acid as predicted by the model. Adaptive laboratory evolution of the chassis-derived producer cells further improved production for succinate and fumarate attesting to the evolutionary robustness of the underlying growth-product coupling. In the case of malate, which exhibited decreased production during evolution, the multi-omics analysis revealed flux bypass at peroxisomal malate dehydrogenase not accounted in the model. Transcriptomics, proteomics and metabolomics analysis showed overall concordance with the flux re-routing predicted by the model. Together, our results provide experimental evidence for model-based design of microbial chassis and have implications for computer-aided design of microbial cell factories.