Human milk is a complex and dynamic biological fluid. Besides the basic provision of nutrition, it contains a wide range of bioactive molecules that support the immune and cognitive development of the infant, as well as their metabolic processes. Understanding the chemical diversity of human milk and quantifying how it differs from common artificial substitutes is thus central to the improvement of infant nutrition. In this work, a chemoinformatic approach has been adopted to characterize and compare the chemical spaces pertaining to human, bovine, and plant-based (i.e. soy) milks, as well as selected weaning foods. This was subsequently complemented with compounds with known bioactivity from curated molecular datasets, collected through a comprehensive review of the scientific literature and from specialized databases, including FooDB, MilkyBase, and PubChem. To represent complementary dimensions of molecular diversity, multiple types of molecular encodings were generated, such as physicochemical descriptors, structural fingerprints, and deep learning-based embeddings such as ChemBERTa. Dimensionality reduction approaches, e.g. PCA, t-SNE, and VAE-derived latent spaces, allowed for the visualization and interpretation of the associated chemical landscapes. Human, cow, and soy milks occupied distinct but partially overlapping areas in chemical space: the former two were closer because of their related composition and biological functions, while soy milk was more distant, as expected for a plant-based surrogate. Weaning foods occupied a larger and less compact area, partially overlapping that of human milk: this may be indicative of a gradual metabolic adaptation in early life. Remarkably, the chemical space of human milk was enriched in the neighborhood of molecules with reported anti-inflammatory, antibacterial, antioxidant, and antiviral activities, suggesting functional convergence. Overall, this work provides a computational framework that forms the basis for the in-silico exploration of the composition of human milk in tight connection with its molecular functionalities. By modeling and visualizing its molecular diversity, it contributes to the creation of data-driven approaches able to simulate and predict the nutritional and bioactive features of milk. It is thus an important step toward the personalization of infant nutrition.