This paper provides a comprehensive review of the state-of-the-art (SOTA) in automated machine learning (AutoML), focusing on deep learning (DL) techniques. It highlights the challenges of building high-quality DL systems, which often rely heavily on human expertise, and introduces AutoML as a promising solution. The paper covers various aspects of the AutoML pipeline, including data preparation, feature engineering, hyperparameter optimization (HPO), and neural architecture search (NAS). A detailed discussion on NAS methods is provided, including one/two-stage NAS, one-shot NAS, joint hyperparameter and architecture optimization, and resource-aware NAS. The performance of representative NAS algorithms on the CIFAR-10 and ImageNet datasets is summarized. The paper also addresses open problems in AutoML and concludes with a discussion on future research directions.This paper provides a comprehensive review of the state-of-the-art (SOTA) in automated machine learning (AutoML), focusing on deep learning (DL) techniques. It highlights the challenges of building high-quality DL systems, which often rely heavily on human expertise, and introduces AutoML as a promising solution. The paper covers various aspects of the AutoML pipeline, including data preparation, feature engineering, hyperparameter optimization (HPO), and neural architecture search (NAS). A detailed discussion on NAS methods is provided, including one/two-stage NAS, one-shot NAS, joint hyperparameter and architecture optimization, and resource-aware NAS. The performance of representative NAS algorithms on the CIFAR-10 and ImageNet datasets is summarized. The paper also addresses open problems in AutoML and concludes with a discussion on future research directions.