Intelligent generative design framework for semantic driven mechanical design

Mechanical product design integrates functional reasoning, geometric modelling, assembly planning, and engineering validation, yet remains highly dependent on human expertise. Existing large language models lack the spatial reasoning and physical constraint awareness required for engineering-grade design, whereas task-specific neural networks perform well in localised geometric recognition but are restricted to narrowly defined and repetitive tasks. Here we show an Intelligent Generative Design framework that enables autonomous mechanical design from natural-language requirements through biologically inspired coordination of task-specific agents. Integrating multimodal engineering knowledge, the framework decomposes design tasks, performs engineering analysis, generates dimensionally accurate three-dimensional models, infers assembly constraints, and validates kinematic behaviour. Across diverse mechanical systems, the framework produces mechanically consistent designs and completes multi-component tasks beyond standalone large language models. Intelligent Generative Design establishes a framework for practical AI-assisted mechanical product development, with implications spanning intelligent manufacturing, rapid prototyping, and design space exploration within diverse industrial domains.