Optimization plays a central role in data-driven design. However, selecting the right optimization algorithm for a particular task is usually not trivial, as that results from the nature of the optimization problem to be solved and constraints such as available resource budget and time to evaluate each data point. In order to address this, there has been a transition from creating hand-designed optimization algorithms to an approach attempting to learn the optimization conditions themselves with machine learning strategies. Although important work has been developed in this field, they often suffer from poor generalization to other distributions, and their knowledge cannot be extended without retraining. We propose a new approach called "Learning to Choose Optimizers" (L2CO). Our method allows a meta-learner to select at test time from a range of static and well-established optimizers, allowing to switch to a different optimizer during the optimization process itself. This enables the system to adapt to different task distributions and enhance generalization performance. The framework is set up to continuously incorporate new optimization algorithms and problems, inviting domain experts to contribute their insights and expertise for ongoing expansion. In this study, we train our model offline on a diverse set of benchmark loss-functions and apply a range of gradient-based, population-based, and probabilistic model-based optimizers. The potential of this approach is demonstrated by comparing its performance to classical optimizers on benchmark loss-functions and simple material design studies. The results suggest that L2CO can be a useful tool and exhibits better generalization. Our open-source code and documentation are available according to FAIR (Findable, Accessible, Interoperable and Reproducible) standards.