This article presents a novel strategy for the non-overlapping disk-packing problem for generating dense random assemblies. The proposed geometric algorithm adopts an advancing-front approach that uses novel heuristics to determine the next positions for incoming particles with support from a polygonal mesh. Additionally, we propose strategies to improve generation at the pack's interior and near container borders. Experiments demonstrate that our algorithm outperforms previous methods in terms of the desired particle radius distribution function, packing density, and mean number of particle contacts.