The deployment pipeline of deep neural networks (DNNs) in safety-critical settings involves formally verifying their trustworthiness. Domain experts design a large number of specifications (~10-100K), typically defined for inputs in the test set, to cover DNN behavior under diverse real-world scenarios. If the DNN is proved to be trustworthy, then it can be deployed. Otherwise, it is repaired or retrained. While there has been a lot of work in recent years on developing precise and scalable DNN verifiers, existing verifiers are inefficient when applied inside the deployment pipeline. This inefficiency is because the verifier needs to be run from scratch for every new specification and DNN.