Human cardiac tissue engineering can fundamentally impact therapeutic discovery through the development of new species-specific screening systems that replicate the biofidelity of three-dimensional native human myocardium, while also enabling a controlled level of biological complexity, and allowing non-destructive longitudinal monitoring of tissue contractile function. Initially, human engineered cardiac tissues (hECT) were created using the entire cell population obtained from directed differentiation of human pluripotent stem cells, which typically yielded less than 50% cardiomyocytes. However, to create reliable predictive models of human myocardium, and to elucidate mechanisms of heterocellular interaction, it is essential to accurately control the biological composition in engineered tissues. To address this limitation, we utilize live cell sorting for the cardiac surface marker SIRPα and the fibroblast marker CD90 to create tissues containing a 3:1 ratio of these cell types, respectively, that are then mixed together and added to a collagen-based matrix solution. Resulting hECTs are, thus, completely defined in both their cellular and extracellular matrix composition. Here we describe the construction of defined hECTs as a model system to understand mechanisms of cell-cell interactions in cell therapies, using an example of human bone marrow-derived mesenchymal stem cells (hMSC) that are currently being used in human clinical trials. The defined tissue composition is imperative to understand how the hMSCs may be interacting with the endogenous cardiac cell types to enhance tissue function. A bioreactor system is also described that simultaneously cultures six hECTs in parallel, permitting more efficient use of the cells after sorting.