Intellegens has announced the latest release of its machine learning software, focused on key practical tasks to accelerate innovation for chemicals, materials, and manufacturing.
Alchemite applies an Artificial Intelligence (AI) method developed at the University of Cambridge to optimise products and processes, reduce experimental workloads, and enable faster R&D progress.
Ben Pellegrini, CEO at Intellegens comments: “We work closely with our chemical, materials, and manufacturing customers to focus our development work. Often, it’s the small details that really make the software work in an industrial R&D environment. We’re delighted that more organisations now benefit from Alchemite. Significant customer agreements so far this year have included a major steel manufacturer, world-leading speciality chemicals providers, additive manufacturing specialists, and a top global food producer."
Highlights of the release are UX and performance enhancements, improved handling of real-world experimental and process datasets, and explainable AI tools that provide greater insights into chemicals, materials, and processes.
User experience advances have focused on streamlining the browser-based Alchemite Analytics platform for key tasks. These include: training new machine learning models; identifying optimal inputs (e.g., chemistries, ingredients, or processing steps) to achieve target properties; and guiding acquisition of new data to ensure experimental efficiency while improving model performance. New users can also get started quickly, with the added ability to train models from very small datasets, improved sign-in workflows, and integrated documentation. Combining these many small enhancements means users get better results, faster.
Alchemite already offers unique capabilities to model real experimental and process data through its unique ability to handle sparse, noisy datasets. The Autumn Release builds on this strength with new tools that ease work with categorical data. Users can now specify and analyse such data using convenient text labels (e.g., colours, process types, or materials classes). Algorithm enhancements have speeded-up model performance for optimisations that include categorical data. Experimental design capabilities now offer greater control, for example, by enabling users to limit the number of ingredients included when Alchemite™ suggests what new experiments to try. This helps to manage costs and, more importantly, to work within practical limitations of real-world experimental programs.
New analytics provide greater insights into chemicals, materials, and processes. They help users to answer questions such as: “which process parameters really affect the viscosity of my formulation?” or “which ingredients should I focus on controlling in order to minimise risk of material failure?”. Dimensionality reduction visualisation enables users to view in two dimensions a graphical representation of their entire, multidimensional dataset in order to gain a deeper understanding of their design space.
