We experiment with a contextual approach to learning monitors for ML-based components. Particularly, in systems where multiple neural-based controllers are available, the monitor’s task is to determine, based on the observed context, which controller is best suited to maintain the system's safety. Our approach utilizes techniques from contextual multi-armed bandits and statistical verification to learn monitors. To validate our approach, we conduct a case study in the domain of autonomous driving which requires simulation in high fidelity simulation environment and the deployment of statistical verification approach for which we require the computation resources.