AlloyGPT: an agent-based LLM framework for the design of additively manufactured structural alloys in extreme environments

Traditionally human are involved in various planning, collecting knowledge, designing, and validating structural alloys for a given application. This webinar includes a presentation of a large language model powered by collaborations of various AI agents to automate this process and accelerate the material discovery. The speaker talks about how a framework was applied for the design of high-temperature strength printable Al alloys with thermal stability. As well as showcase how GPT-based LLMs which are trained by CALPHAD-based ICME data can predict microstructural features and properties in comparison with Bayesian optimization and conventional machine learning techniques.

Furthermore, there is a discussion of the accuracy of the model in forward prediction and inverse design, its unique opportunities, and the efficiency in combining different agents in this design. This hybrid framework can lay the foundation for the automatic design of structural alloys which are manufactured by various techniques and are in extreme environments.