<p>To design a lipid nanoparticle (LNP) that effectively delivers nucleic acids to a specific cell or tissue type, multiple lipid components and their relative proportions must be decided on from a large number of options. As there is an incomplete understanding of the relationship between the molecular composition of a delivery vehicle, its structure and its activity, the decision is made by screening many formulations. Emerging technologies have rapidly accelerated the generation of large LNP libraries and the testing of their physicochemical properties and behaviour in vitro and in vivo. These screening tools are being increasingly integrated within artificial intelligence-driven discovery systems, wherein data obtained from the characterization and biological testing of LNPs are fed into machine learning models. These models can provide non-obvious relationships between composition and physical or biological outputs, or predict entirely new lipid structures. In this Perspective, we discuss advancements in the automation and parallelization of chemical synthesis, particle formulation, characterization and pharmacological screening that have improved the throughput of generating and testing large libraries of LNPs for nucleic acid delivery. We notably highlight the short-term potential of coupling these high-throughput platforms with machine learning to accelerate the prediction of optimal nucleic acid LNPs for new therapeutic targets.</p>