In modern applications, complex and high dimensional data are collected, which can be used to estimate reliability more precisely. The existing reliability approaches, however, cannot efficiently model the complex effect of covariates on failure time. We propose a novel deep learning-based reliability approach to model the complex relationship between covariates and failure times. To estimate model parameters, neither the traditional deep learning parameter estimation method nor maximum likelihood estimation method is applicable. To overcome the difficulty, a new model parameter estimation method is developed based on partial likelihood framework. Simulation and a real-world case study are conducted to verify the developed methods.