Commercially available engineering alloys are often strengthened through precipitation hardening. Materials are heat treated to induce the formation of a fine distribution of nanoscale precipitates embedded within a matrix phase. Alloy designers carefully tune chemical compositions, aging times and temperatures to improve alloy properties. Optimizing the properties of these materials necessitates a fundamental understanding of the nucleation process. Atomistic simulations are an invaluable tool to elucidate precipitation mechanisms and compute nucleation rates. However, as the length and time scales of nucleation are out of reach for fully ab-initio methods, machine-learned interatomic potentials must be employed to bridge the gap between electronic structure calculations and phenomenological models of nucleation. Recent years have witnessed several advances in the parameterization of machine-learned interatomic potentials. The development of effective training datasets and validation methodologies for metallurgical alloys remains an ongoing challenge. This talk will outline several key considerations in the parameterization of multicomponent interatomic potentials for studying precipitation that are often overlooked. The work demonstrates that carefully designed training datasets, together with novel ways of quantifying the precision of the interatomic potential that look beyond conventional statistical errors, are necessary to attain the degree of accuracy necessary to quantitatively reproduce the phase transformations occurring during precipitation. The study of a magnesium-rare earth alloy illustrates the development of a machine-learned interatomic potential to capture complex precipitation sequences, order-disorder phase transitions, and structural transitions. Finally, atomistic simulations with an accurate machine-learned interatomic potential quantify the thermodynamic and kinetic barriers for nucleation in the magnesium alloy. The results and methods outlined in this talk serve as important guidelines for parameterizing accurate potentials for metallic alloys and provide key insights into solid-state nucleation mechanisms.