Machine learning algorithms need data, so the researchers designed a polymerization process that would quickly and efficiently generate experimental data to feed into the mathematical model.