Functional grading is a distinctive feature adopted by nature to improve the transition between tissues that present a strong mismatch in mechanical properties, a relevant example being the tendon-to-bone attachment. Recent progress in multi-material additive manufacturing now allows for the design and fabrication of bioinspired functionally graded soft-to-hard composites. Nevertheless, this emerging technology depends on several design variables, including both material and mechanistic ingredients, that are likely to affect the mechanical performance of such composites. In this paper, a model-based approach is developed to describe the interaction of ultrasound waves with homogeneous and heterogeneous additively manufactured samples, which respectively display a variation either of the material ingredients (e.g., ratio of the elementary constituents) or of their spatial arrangement (e.g., functional gradients, damage). Measurements are performed using longitudinal bulk waves, which are launched and detected using a linear transducer array. First, model is calibrated by exploiting the signals measured on the homogeneous samples, which allow identifying relationships between the model parameters and the material composition. Second, the model is validated by comparing the signals measured on the heterogeneous samples with those predicted numerically. Overall, the reported results pave the way for characterizing and optimizing multi-material systems that display complex bioinspired features.