The creation of civil structures such as bridges, dams, and buildings involves a collaborative effort among various experts, each handling specific aspects of the process. Traditionally, this process is performed sequentially, i.e. starting with the structural design with assumptions on minimal material properties, and a subsequent material design within these limits. In the presentation, an alternative procedure is proposed that merges concrete mix design and structural simulations into a unified, progressive workflow. The design procedure is then reformulated as an optimization problem with constraints related to key performance indicators such as e.g. the load bearing capacity. Addressing uncertainties—whether inherent (aleatoric) or arising from incomplete knowledge (epistemic)—is crucial in any design endeavor. These uncertainties may stem from data used to calibrate physical models, from model simplifications or from machine learning models trained with limited amount of data that are used to bridge gaps in the workflow. Inverting the causal relations poses several challenges especially when these physics-based models do not provide derivatives/sensitivities or when design constraints are present. Our approach advocates for Variational Optimization, augmented by proposed extensions and carefully selected heuristics. Integrating the expertise and numerical methods from multiple experts is a challenging task in itself. The presentation discusses the use of workflow tools with a modularization of the computational methods. The presentation includes related challenges in the context of the FAIR principles (Findable, Accessible, Interoperable, Reusable). The design of a precast concrete beam is presented as an example to demonstrate the efficacy of our methodology. The objective is to minimize the environmental impact in terms of the global warming potential while fulfilling constraints related to the load-bearing capacity (Eurocode) as well as the demoulding time and the maximum temperature during hydration.The latter are computed from a macroscopic thermo-mechanical Finite Element simulation that is based on a hydration model on the paste level combined with homogenization procedures to obtain the macroscopic constitutive equations. Experimental data is used to perform a Bayesian calibration of these models.