Pipit: A foundation model factory for manufacturing and materials engineering using agentic code generation
This project seeks to develop Pipit, an AI-based “foundation model factory” that will create custom foundation AI models that U.S. industry can leverage to accelerate materials development and the application of advanced manufacturing methods such as 3D printing. Foundation models hold the potential to revolutionize how insights and learned relationships are implemented across industrial enterprises, but realizing this potential requires extensive data generation and curation, AI model construction and training, and deployment on large-scale high-performance computing architectures. Pipit will provide an AI-enabled infrastructure that uses coding agents to convert prototype machine learning models into end-to-end foundation model pipelines. Pipit’s foundation models, benchmarks and data generated will be shared on the Genesis Mission platform for community use.
Partners
- University of Texas, Austin
- GE Vernova