Polymer informatics is an advanced field enabling the prediction and inverse design of polymer properties through machine learning, employing descriptors that numerically encode multi-layered information about polymer structures. This article provides an introductory overview of polymer descriptors. It discusses monomer-level descriptors, methods involving optical and electrical properties derived from Density Functional Theory (DFT) calculations, extraction of force-field parameters from Molecular Dynamics (MD) simulations as descriptors, and techniques utilizing macroscopic characteristics like thermal conductivity and elastic modulus obtained from MD simulations. The integrated use of these multi-scale descriptors is expected to enhance prediction accuracy and interpretability, leading to new insights into the relationship between molecular behaviour and polymer properties, as well as enabling sustainable material development.