Dr. Taline Kerackian - Automation and Machine Learning for Electrosynthesis

Organic electrosynthesis comes with specific variables in addition to those found in thermochemical reactions, that will require optimization such as supporting electrolyte nature and concentration, current density, amount of applied charge, electrode material and interelectrode distance. Reaction optimization is a central phase of chemical synthesis. Despite being material- and time-intensive, this process is critical for converting an initial hit into a high-yielding reaction. Our group is interested in making the optimization process more efficient and economical. One Variable at a Time is the most common optimization technique used in synthetic chemistry. Our group has widely used Design of Experiments as a statistical optimisation method for various electrosynthesis transformations. But, to further reduce the time and effort for reaction optimization, Bayesian Optimization has emerged. By exploiting known reaction outcomes, it can explore the parameter space in a balanced way to find a global optimum. We want to exploit Artificial Intelligence and Machine Learning potential to help us tackle electrosynthesis challenges.

However, Bayesian Optimization requires high quality data which are reliable, reproducible and standardized. Automation and Self-Driving Labs have provided such results. Automating all synthetic steps required to perform a set of chemical reactions allows to minimize the human related uncertainty. We aim at pushing the boundaries of what automation can bring to electrosynthesis.