Machine learning-driven agrochemical formulation design and optimization platform

Overview

Agrochemical formulation development traditionally requires extensive trial-and-error experimentation to balance competing performance objectives such as active ingredient stability, biological activity, and cost. This platform applies machine learning to model the complex physicochemical interactions between formulation components, enabling faster and more efficient design of organic agrochemical formulations. By representing molecular function and component interactions within a hierarchical machine learning framework, the technology predicts formulation performance and screens new chemical species for their potential role in complex mixtures. The approach reduces development time and material costs while expanding the design space available to formulation chemists.

Technical specifications

Core capabilities:

• Hierarchical machine learning models that integrate chemical structure representations with known physicochemical properties of formulation constituents
• Prediction of formulation performance across multiple objectives including stability, activity, and cost optimization
• Virtual screening engine for rapidly evaluating new components as candidates for complex formulations
• Integration of physicochemical and engineering models within a unified machine learning framework
• Adaptability to diverse active ingredients and formulation types

Proven track record:

The underlying methodology has been validated with major chemical industry clients including PPG, Evonik, Lubrizol, DIC, and Covestro in non-agricultural formulation applications, demonstrating broad applicability across formulation chemistry domains.

Technology readiness level

The machine learning methodology is validated and operational in industrial formulation design contexts. Current development focuses on extending the platform specifically to agrochemical formulations through a data partnership with Valent, where formulation data for diverse active ingredients will be used to build and refine domain-specific predictive models. The technology is positioned for collaborative development with agrochemical industry partners seeking to accelerate product development pipelines.