Structures of compounds found in chemical libraries are inputted into our AI foundation model. Based on chemical structure, the foundation model generates predictions for these compounds that relate to
their bioactivity and toxicity.

Compounds that are predicted to meet a certain threshold of bioactivity and safety are sent to another AI module that screens the compounds for commercial feasibility in terms of manufacturability and intellectual property.

Compounds that pass are screened on the company's robotic high-throughput cell screening platform. The multimodal data produced by high-throughput cell screening are fed back to the foundation model to
continually improve its accuracy.

The most promising compounds are then tested on our long-lasting models of human skin, and the multimodal data produced by these assays are fed back into the foundation model.

This machine-learning workflow allows us to rapidly assess bioactivity across multiple skin health applications.

Our foundation model can computationally pre-screen approximately 750 million compounds a year, increasing speed and efficiency of novel active compounds discovery.