Application domains of Bayesian optimization include optimizing black-box functions or very complex functions. The functions we are interested in describe complex real-world systems applied in industrial settings. Even though they do have explicit representations, standard optimization techniques fail to provide validated solutions and correctness guarantees for them. In this paper we present a combination of Bayesian optimisation and SMT-based constraint solving to achieve safe and stable solutions with optimality guarantees.