Vector-borne diseases are infectious diseases transmitted between humans by an animal vector, e.g. a mosquito a tick. Globally, one million people are killed by vector-borne disease each year. Current interventions are limited in their efficacy, applicability, and safety profile. In collaboration with the Gates Foundation and Google we set out to discover novel candidate molecules using a multidisciplinary approach that leveraged machine learning, computer vision, and advanced laboratory automation. We digitized thousands of records of repellency and acaricide activity for training of predictive graph neural networks (GNNs). Model predictions were verified in automated bioassays against a variety of arthropods and through live imaging of mosquito brain activity. As a result, we established a computational representation for odor at the molecular level that can predict behavioral responses in a wide variety of organisms. This approach allowed us to identify more than 100 highly effective arthropod repellents, including compounds outperforming market standards in human volunteer studies. Live imaging of calcium signaling in mosquito brains revealed distinct signaling pathways for structurally diverse compounds. Computer vision-based analysis of tick movement provided insights into the mechanisms of action of repellents, distinguishing between contact irritants and spatial repellents. In an adaptation of the approach, we trained GNNs on acaracide data. Robotic arms and custom-designed 3D-printed components enabled high-throughput assays to evaluate predicted acaricidal activity following contact or systemic exposure. The results revealed a 6% hit rate on predicted acaricides from a chemical diversity screen. These compounds form the basis for further refinement of the GNN models and hit optimization. The combined data demonstrate how a data-driven approach can significantly enhance the speed and efficiency of small molecule discovery, leading to accelerated progress towards solving a global health challenge.