MaterialsGPT: A Domain-Specialized Language Model for Accelerating Materials Discovery

Machine learning (ML) is transforming materials research by shifting from intuition-driven and trial-and-error approaches to data-driven discovery.Researchers are increasingly leveraging AI/ML methods to guide experiments and accelerate innovation. Recent advancements in large language models (LLMs) have demonstrated strong capabilities in summarization, reasoning, and analysis. However, existing LLMs lack domain-specific understanding of materials science, limiting their applicability to materials discovery.

This project aims to develop MatterialsGPT—an accessible, costeffective, and robust domain-specialized LLM designed to extract materials knowledge from scientific literature, uncover hidden structure-property relationships, guide experimentalists, and accelerate the discovery of novel materials. Through this work, the candidate will gain hands-on experience at the intersection of AI and materials science while contributing to the development of an AI-driven tool for materials research.

MatterialsGPT has the potential to drive discoveries in various fields including flexible electronics, photovoltaics, energy storage, and other emerging fields, and is aligned with several UN sustainability goals.