Nowadays, to meet the vision of employing 100% renewable-based electricity generation, the conventional power system is evolving into power electronics-dominated grid (PEDG). This transition leads to an amplified complexity and significance for device and system-level control schemes to maintain resiliency, reliability, and operational stability. Recently, in various fields of engineering and science, the machine learning (ML)-based schemes have exhibited outstanding performance. Considering abundance of data in the PEDG, ML- based approaches illustrate promising potential to be adopted in this new energy paradigm. Similarly, the ML- inspired approaches have been attracting many researchers in power electronics and power systems fields, who are trying to solve the challenges posed by the PEDG concept. This article presents cutting-edge ML applications in the PEDG and provides a futuristic research roadmap.