Co-amorphous systems (CAMs) have shown promise in addressing the challenges associated with poorly water-soluble drugs. However, the limited selection of co-formers and the use of lab-scale techniques for their preparation present challenges in fully utilizing the advantages of CAMs. In this study, we used aspartame (a methyl ester of the aspartic acid/phenylalanine) as a model dipeptide with the BCS class II drug dipyridamole, to prepare co-amorphous systems using spray drying. The feed solutions were prepared by dissolving the drug and co-former into methanol–water mixtures. The spray drying process was evaluated and solid-state properties were compared with those of the individual amino acids, amino acid mixtures and aspartame as co-formers. Co-amorphous systems prepared with aspartame (AspPhe) exhibited better solid-state properties, including a higher glass transition temperature (Tg), compared to the individual amino acids and the mixture of amino acids. Additionally, this formulation showed improved physical stability when stored at 25 °C/60 % RH conditions. Hirshfeld Surface (HS) analysis was employed to visualize and analyse the molecular interaction sites within the crystal structures of dipyridamole and aspartame. The observed interactions were then correlated with the molecular interactions identified through FT-IR spectroscopic analysis within the CAMs. The spectroscopic analysis revealed molecular interactions between the sites found at the shortest distances in the HS analysis. The dominant hydrogen bond interactions identified in the co-amorphous DPM-AspPhe system was found to contribute significantly to its improve stability. X-ray powder diffraction in non-ambient mode reveals that both temperature and humidity play a role in the crystallization of the co-amorphous DPM-AspPhe. Crystallization rates increased notably at high temperature and humidity. To predict stability under accelerated conditions, the crystallization rates from DPM-AspPhe were fitted to a modified Arrhenius equation. However, the predictive accuracy of the resulting model was limited to a specific range of conditions.