Reliability estimation is essential for the design and operation of engineered systems. Calculating the reliability of large k-out of-n systems becomes computationally intensive as it involves summing numerous combinations of functioning and failed components. Direct enumeration methods can be used, but their runtime grows quickly, limiting their usefulness for repeated analysis. This study investigates artificial neural networks (ANN) as surrogate models to provide fast and accurate reliability estimates of a k-out of-n system. A large synthetic dataset of components’ reliability is generated that is used for the training and evaluation of the neural network model. A multi-layer perceptron model is then trained to learn the nonlinear mapping from components’ reliability to overall system reliability, with the output constrained to the [0,1] interval. This use of ANN is assessed using standard regression metrics and a deviance-based metric. Its computational results are compared against direct analytical computation across multiple k-out of-n settings. In addition, runtime is evaluated by comparing neural-network inference against direct analytical computation, showing that the proposed approach can deliver faster reliability estimates for larger k-out of-n systems. The ANN approach helps in estimating the reliability of k-out of-n systems with reasonable accuracy at much faster computational times.