A single image containing two separate photos of two men smiling at the camera.

CMU Startup Helps Formulators with Complex Design Problems

By Heidi Opdyke Email Heidi Opdyke

Heidi Opdyke

Working with development partner Procter & Gamble, Carnegie Mellon University startup Ansatz AI recently announced the release of Hierarchical Machine Learning software. Designed for personal care and consumer products, this powerful tool makes design and optimization of a diversity of formulated products simple.

The startup was founded by Newell Washburn, associate professor of chemistry and biomedical engineering with a courtesy appointment in the department of materials science and engineering; and Barnabás Póczos, associate professor of machine learning.

“I was always interested in designing molecules and materials to have a certain function, and wondered if we could use machine learning to do so if we didn’t have a lot of data,” Washburn said.

Washburn and Póczos discussed how to use machine learning to analyze sample sizes as small as 20 to 30 items rather than samples sizes of hundreds or thousands. Along with the help of students, they built their first algorithm, which they called Hierarchical Machine Learning, HML is a data processing system for analyzing data records in designing a formulation of a material. The system’s development was supported by an NSF grant.

Through resources in Carnegie Mellon’s entrepreneurial ecosystem, the researchers patented their work and licensed it through the Center for Technology Transfer and Enterprise Center (CTTEC). The center guides faculty and students through the process of transferring invention to industry where they can be developed into commercial products. Washburn also consulted with contacts in Carnegie Mellon’s Project Olympus, which offers faculty and students startup advice, micro-grants, incubator space, and connections to turn research and ideas into businesses.

Ansatz AI began as a consulting firm and has worked with companies in the United States, Europe and Japan on problems in molecular design, formulation and process optimization. Ph.D. student, Ansatz co-founder and chief system designer Calvin Gang took that process and built an interactive design tool.

“The critical piece for us is that we ask our clients a lot of questions about how something works and what chemical interactions control results, and how you model that,” Washburn said.

Ansatz AI’s platform helps clients by providing a systematic framework for understanding how various compounds may work together based on their properties. Once the platform has been provided candidate chemical forces and interactions, it builds a model for how the product works. Based on that model, formulators can virtually experiment by adjusting the amounts and types of compounds before identifying a limited number of formulations to test in real life.

Carnegie Mellon and its startup companies like Ansatz are committed to engaging with U.S. industry in solving a broad range of technological challenges.