From Data to Innovation: How Premix is Using AI to Accelerate Recipe Development

Virtual, AI assisted formula development is gaining momentum across both industry and academia — and for good reason. By leveraging theoretical models and machine learning, companies can speed up development, reduce iteration cycles, and optimize use of resources. 

At Premix, we’ve built on our extensive knowledge of functional plastics to create a powerful AI-assisted tool: PERTTI 3.0 — a machine learning–driven system that enhances how we develop and refine our electrically conductive compound formulations. 

What is PERTTI 3.0? 

Developed in-house by our R&D team, PERTTI 3.0 (short for Premix 'Recipe' Tool for Technical Insights) is an AI-enabled system trained on more than four decades’ worth of formulation data — including thousands of compound recipes/ formulas developed. 

The tool currently supports three core functionalities: 

• Data printing: Instantly retrieves and displays key material properties for any given recipe/formula, even for those raw materials that have never been physically tested earlier. 

• Recipe ranking: Analyzes existing compounds to identify best-fit solutions, reducing duplicate development efforts. 

• Recipe generation: Uses predictive modeling to simulate thousands of new randomly generated formulations. Results are filtered based on target properties (e.g., tensile modulus, impact strength) and optimized for criteria like cost or processing requirements. 

“Machine learning itself isn’t new,” says Ville Mylläri, Product Development Manager of Premix. “What’s changed is the accessibility. With the rise of open-source tools, even specialized manufacturers like us can now build powerful AI models with relatively low cost and high flexibility.” 

Designed by Experts, Built for Flexibility 

PERTTI 3.0 is built using open-source frameworks and a modular architecture, which makes it easily adaptable across different polymer types and use cases. This flexibility is key as Premix continues to diversify its material portfolio — especially with the expansion of manufacturing in the U.S. and growing demand for tailored solutions in new markets. 

“Our next steps involve expanding the supported polymer families and scaling the model to support our new regions,” Mylläri explains. “The goal is not to replace human expertise, but to give our product developers faster, smarter tools to deliver better outcomes for our customers.” 

What This Means for Our Customers 

With AI-powered support from PERTTI 3.0, Premix can: 

• Accelerate formulation screening and iteration 

• Deliver faster time-to-sample 

• Support more informed material selection 

• Reduce waste and redundancy in product development 

In short, customers benefit from more agile development cycles, higher confidence in early-stage recipes, and faster delivery of customized materials — backed by science, data, and decades of compounding know-how.