Domain Knowledge Integration
Make better predictions and uncover novel insights with less data. Encode your scientific, production, and business knowledge into the Citrine Platform so that AI models build on your knowledge instead of recreating it.
Creation of a Domain Knowledge-Enabled Workflow
Product developers, data managers, and data scientists can work together using the Citrine Platform to integrate domain knowledge throughout the AI workflow.
Incorporating domain knowledge starts with creating a data structure that captures the full context of the material’s process history and properties. The platform also facilitates the conversion of specialist data, such as chemical formulas, into related machine-readable data, such as molecular weight, via 100+ available “featurizers”. Customers can also build their own featurizers.
Graphical, no-code AI models, enable researchers to understand and contribute to AI model building – making models more accurate. Known laws of physics do not have to be reinvented, they can be incorporated using mathematical expressions. Knowledge of intermediate properties and mechanisms can be built in as latent variables.
Product developers can use their understanding of what the market needs and any constraints that are in play to set up the search space the AI Model should explore. Ingredient compatibility, availability and cost, and processing parameter constraints can all be used to define a search space that will produce feasible candidates.
Property targets can be set based on the knowledge of what the customer needs. The search strategy can be chosen depending on the time scale of the project.
Actionable IP Assets
The resulting data, search spaces, and models constitute an actionable IP asset that can be shared with peers and used to power future projects which share chemistry, ingredients, or processing.