In this talk, I will discuss how algorithms for neural network verification can be used to solve robotics problems. I will first introduce our toolbox NeuralVerification.jl, followed by the discussion of two applications. The first is to build verified models for human behavior prediction in human-robot systems. The second is to develop safe controllers for dynamic models encoded in deep neural networks. To support real-time computation in robotics tasks, I will also discuss potential approaches to speed up the computation of the verification algorithms.