Various simulation techniques offer different levels of accuracy and scalability. Generally, to achieve higher accuracy (or scalability), one has to compromise on scalability (or accuracy). One such technique, molecular dynamics, offers a balance of accuracy and scalability by relying on the interatomic potential derived from experimental and density functional theory (DFT)-based results. However, the advancement of machine learning (ML) in materials science has led to the development of data-driven interatomic potentials, which significantly enhance the accuracy and scalability of simulations. This chapter focuses on the Spectral Neighbor Analysis Potential (SNAP), which is an ML-based interatomic potential designed to accurately and efficiently model atomic interactions in complex materials. SNAP utilizes bispectrum components derived from hyperspherical harmonics to represent local atomic environments and employs linear regression to map these descriptors to energy and forces. SNAP also ensures key invariances being invariant under translation, rotation, and permutation of identical atoms, while offering significantly improved computational efficiency compared to earlier ML potentials like the Gaussian Approximation Potential (GAP). This chapter outlines the theoretical framework behind the SNAP, including its descriptor construction and regression model. Applications across various material systems have also been reviewed to highlight the SNAP predictive capability. In brief, the SNAP offers a robust framework for precise, large-scale atomistic simulations, allowing for the investigation of material phenomena that were previously inaccessible using conventional methods.