Deep neural networks have seen great success in recent years; however,\ntraining a deep model is often challenging as its performance heavily depends\non the hyper-parameters used. In addition, finding the optimal hyper-parameter\nconfiguration, even with state-of-the-art (SOTA) hyper-parameter optimization\n(HPO) algorithms, can be time-consuming, requiring multiple training runs over\nthe entire dataset for different possible sets of hyper-parameters. Our central\ninsight is that using an informative subset of the dataset for model training\nruns involved in hyper-parameter optimization, allows us to find the optimal\nhyper-parameter configuration significantly faster. In this work, we propose\nAUTOMATA, a gradient-based subset selection framework for hyper-parameter\ntuning. We empirically evaluate the effectiveness of AUTOMATA in\nhyper-parameter tuning through several experiments on real-world datasets in\nthe text, vision, and tabular domains. Our experiments show that using\ngradient-based data subsets for hyper-parameter tuning achieves significantly\nfaster turnaround times and speedups of 3$\\times$-30$\\times$ while achieving\ncomparable performance to the hyper-parameters found using the entire dataset.\n