Description
Artificial intelligence is rapidly changing how materials research is conducted, not only through machine learning models but also through emerging AI-assisted research practices such as vibe coding, where natural-language interaction is used to develop data-analysis, visualisation and modelling tools. This work presents a human-in-the-loop AI workflow developed to support the design and optimisation of hybrid graphene-based nanocomposites containing Polyamide 6 (PA6) reinforced with graphite, graphene nanoplatelets (GNP) and multi-walled carbon nanotubes (MWCNT). The approach combines AI-assisted coding, machine learning and experimental validation within a single research framework.
Experimental datasets were combined with interpretable predictive models to explore composition–property relationships, identify promising formulation windows and support experimental decision-making. The resulting tools were integrated into an AI-assisted workbench capable of screening candidate materials, exploring composition space and prioritising high-value experiments for validation. Rather than replacing laboratory testing, the workflow was structured as an iterative prediction–verification loop in which model outputs informed experiments and experimental results refined subsequent models.
The study demonstrates how AI can function as a practical research collaborator, accelerating coding, data analysis and scientific decision-making while maintaining human oversight and physical interpretability. More broadly, it highlights the potential of AI-assisted workflows, materials informatics and human-in-the-loop modelling to support future materials discovery and optimisation.