Abstract: As semiconductor manufacturing progresses toward the atomic scale, nanodevices increasingly demand diverse materials and ultra-precise deposition control. Atomic Layer Deposition (ALD) and Atomic Layer Etching (ALE), as essential atomic-scale fabrication techniques, face growing challenges in optimizing high-dimensional and complex process parameters. Traditional simulations and experimental methods often fall short in modeling intricate reactions or supporting high-throughput optimization, highlighting the need for integrated innovations across computational materials science, data science, and artificial intelligence. This work reviews recent advances in applying machine learning to key tasks in atomic manufacturing, including precursor selection, reaction pathway prediction, process parameter modeling, control optimization, molecular dynamics simulations, and data structuring. Machine learning has shown great promise in boosting modeling efficiency, improving predictive accuracy, and enabling adaptive process control. However, challenges remain, such as limited generalization across systems and reduced prediction accuracy under sparse data. Looking forward, combining machine learning with physical constraints, multiscale modeling, and semantic data frameworks may pave the way for a transition from offline prediction to intelligent closed-loop control in next-generation atomic manufacturing.