(101b) Overcoming Challenges in AI for Formulations Development

A distinctive challenge of formulations development is the vast spectrum of possible ingredient combinations. This challenge is intensified by complex rules governing chemical compatibility, processing requirements, and health and safety. From a mathematical perspective, exhaustively exploring this space is difficult due to its high dimensionality and complex boundaries between valid and invalid formulations. From an engineering perspective, this often results in hesitancy to run experiments that are dissimilar from proven solutions.

Here, we will discuss how Citrine Informatics addresses these challenges using tools specifically designed with formulations in mind. These include a graphical data structure for capturing hierarchical information [1], formulations-specific featurizations [2], optimized strategies for considering a wide variety of candidate formulations, and uncertainty-driven strategies for ranking proposed candidates [2]. We will discuss real-world examples of how the candidates proposed by this suite can guide experts to novel fit-for-purpose formulations in significantly less time.

[1] https://citrineinformatics.github.io/gemd-docs/

[2] https://citrineinformatics.github.io/citrine-python/formulations_example.html