Show HN: Autonomous CAD design and OpenFOAM optimization loop using local LLMs (github.com)

A new autonomous multi-agent system has been unveiled for the parametric design, simulation, and optimization of 3D-printable quadcopter propellers, utilizing local large language models (LLMs) alongside tools like CadQuery and OpenFOAM for computational fluid dynamics (CFD). The project aims to automate the traditionally tedious process of tweaking propeller designs, balancing competing objectives such as noise reduction, efficiency, and thrust—challenges that often require extensive trial and error from engineers. By leveraging AI agents to suggest designs, evaluate performance using trusted physics calculations, and refine results iteratively, the system seeks to find optimal trade-offs much faster than manual efforts. This project is particularly significant for the AI/ML community as it advances the concept of AI as a productive worker rather than just a conversational agent. The architecture features a clear separation of roles among various AI models, with a deterministic genetic algorithm ensuring process continuity and reliability. The system includes structured outputs and validation, allowing for a robust design loop that improves over time. Notably, the agents generate new candidate designs while a dedicated physics engine provides unbiased scoring, showcasing a practical application of AI that fosters trust and accuracy in engineering tasks. This methodology could inspire the automation of design processes in other domains, highlighting the adaptable nature of these AI-driven workflows.