Bayesian optimization with known experimental and design constraints for chemistry applications

Optimization strategies driven by machine learning, such as Bayesian optimization, are being explored across experimental sciences an efficient alternative to traditional design of experiment. When combined with automated laboratory hardware and high-performance computing, these enable next-generation platforms for autonomous experimentation. However, the practical application approaches is hampered a lack flexible software algorithms tailored unique requirements chemical research. One aspect pervasive presence constraints in conditions when optimizing processes or protocols, space that accessible designing functional molecules materials. Although many known priori, they can be interdependent, non-linear, result non-compact optimization domains. In this work, we extend our experiment planning Phoenics Gryffin handle arbitrary via intuitive interface. We benchmark extended on continuous discrete test functions diverse set constraints, demonstrating their flexibility robustness. addition, illustrate utility two simulated research scenarios: synthesis o-xylenyl Buckminsterfullerene adducts under constrained flow conditions, redox active batteries synthetic accessibility constraints. The tools developed constitute simple, yet versatile strategy model-based contributing its applicability core component scientific discovery.