How should I efficiently monitor ML models in production?

Question

I have two Neural Network Models that are already deployed. As ML models tend to degrade over time, I need to monitor the NN models in production in terms of efficiency, acc