Single-cell transcriptomics has emerged as a powerful technology to characterise gene expression within individual cells. Over the last 5 years, as the technology has matured, single-cell techniques have been used to profile cells in species across the tree of life. The availability of single-cell data from a rich diversity of species presents an opportunity to compare gene expression, cell types and developmental programmes across organisms and interpret these transcriptomic observations within a phylogenetic framework. This opens up new possibilities to study the evolution of cell types and infer molecular properties of organisms inaccessible to experimental research. In this thesis, I explore these two applications in the context of early mammalian development and vertebrate brain evolution. I first detail efforts to construct a single-cell atlas of rabbit gastrulation, which offers a new window into human embryogenesis and provides a resource for validating therapeutic targets, or observations from mice, in a non-rodent system. I then present three new approaches to visualise and compare single-cell transcriptomics data from multiple species. These methods overcome challenges of relating gene features and cell types, which may be applicable to contexts beyond the cross-species setting. Finally, I apply comparative single-cell analyses to trace networks of neural development across deuterostomes. This reveals a conserved molecular signature present in the sea urchin, amphioxus and zebrafish embryo and offers a template for future studies to compare vertebrate and invertebrate nervous systems with single-cell approaches.