New AI framework AutoSAM automates the traditionally manual process of generating input files for nuclear reactor thermal-hydraulics simulations

AutoSAM combines a large language model agent with retrieval-augmented generation to eliminate labor-intensive manual work in System Analysis Module (SAM) code preparation

The framework processes heterogeneous unstructured engineering documents including system diagrams, design reports, spreadsheets, and PDFs to extract simulation parameters

Multi-modal capabilities allow the system to analyze images, text, and tabular data across multiple document formats automatically

Generated input files are validated and solver-compatible, with results presented in a human-auditable intermediate representation for verification before use