At UNSW’s AI4Science group and our spin out company GreenDynamics, we are pioneering a transformative approach to materials research by fusing advanced AI techniques with physical sciences. Our work redefines how knowledge is extracted from scientific literature and applied to experimental discovery—shifting from a slow, linear process to a scalable and intelligent paradigm.

We have developed AI systems that automate the extraction of structured material-property relationships directly from papers, enabling rapid construction of expert datasets. Using large language models (LLMs), we’ve also demonstrated how to perform scientific question answering, generate hypotheses, and assist in material design and optimization.

This work has led to applications in photovoltaics, self-cleaning coatings, and thin-film materials, where AI not only accelerates insights but also proposes innovative research directions. With tools like SciQAG, DARWIN, and MatFusion, we’re charting a path from “token to discovery,” making materials science more data-driven, reproducible, and exploratory.

Our vision is to unlock the power of foundational AI models for autonomous scientific discovery—moving from perception to action, and from data to innovation.

Learn more: https://unswai4s.notion.site