High throughput experimental study of multi-component phase separating systems is limited by time consuming generation of samples, their measurement, and interpretation. Water-soluble polymers will phase separate dependent on relative concentration, temperature, and composition. Biological phase separation is powerful tool used by nature to mediate intracellular chemistry, survive extreme conditions, and may play a crucial role in climate resilience. Due to partitioning into any phase by any component, applied work is constrained in a two or three component space. We use a high-throughput robotic pipeline to generate thousands of multi-component biopolymer mixtures. Samples are then examined with traditional and ML-based computer vision to examine interfacial, optical, and partitioning behavior using differential interference contrast, confocal fluorescence, and confocal Raman microscopy. This enables a characterization pipeline with size, shape, and chemical information. We use experimental measurement on polymer mixtures to infer structure-property relationships by constraining our pipeline sample space to select optimized mixtures. Finally, these systems are tested in their real world application, providing a final test for inferred behavior. In this talk, the integration of high-throughput experimentation, vision-based characterization and testing will be used to discuss and examine the successes and limitations of using endpoint structure to infer active matter processes and experimental measurement.