Despite its potential to transform society, innovation suffers from a major drawback: its long research timeline with high cost which as well creates a hardle woth the sutainability goals of the industry and the country. Recently, machine-learning techniques have emerged as a viable solution to this drawback. In the competitive world of consumer goods, staying ahead requires constant innovation in formulation development. Companies must navigate rapidly changing consumer preferences, heightened demands for sustainability, and stringent regulatory requirements. There is a need to have a scientific AI governed resolution to address these challenges, streamlining the creation, testing, and optimization of formulations. Whether developing new products or refining existing ones, Scientific AI equips you with the tools needed to maintain a competitive edge in the market. Large Language Models (LLMs) have recently received significant interest from researchers in chemical and materials science.[1,2] However, using LLMs in scientific pipelines presents a challenge for general models due to the specialized terminology and diverse reporting formats used in scientific domains. In this work, we demonstrate several approaches for using LLMs within to perform scientific tasks. We show how machine learning approaches can leverage experimental data to support formulation experts. We will also present results using LLMs to extract and parse data stored in the electronic lab notebooks (ELNs). Overall, our findings highlight the potential for LLMs to aid scientists in the extraction and structuring of data and predictive model development.
References
[1] Z. Xie, X. Evangelopoulos, Ö. H. Omar, A. Troisi, A. I. Cooper, L. Chen, Fine-tuning GPT-3 for machine learning electronic and functional properties of organic molecules, Chem Sci 2023, 15, 500–510.
[2] K. M. Jablonka, P. Schwaller, A. Ortega-Guerrero, B. Smit, Leveraging large language models for predictive chemistry, Nat Mach Intell 2024, 6, 161–169.