Engineering.ai: A Platform for Teams of AI Engineers in Computational Design

TL;DR

Engineering.ai addresses the inefficiencies of multidisciplinary CAE workflows by introducing a hierarchical, multi-agent platform in which a Chief Engineer coordinates Aerodynamics, Structural, Acoustic, and Optimization AI engineers powered by LLMs. The framework enables file-mediated data exchange and memory-based knowledge management to sustain cross-domain collaboration and reproducibility, achieving end-to-end automation demonstrated on UAV wing optimization. The case study shows a twelve-case aerodynamic analysis, a 432-configuration CAD-to-FEA sweep, and GP-Bayesian optimization yielding an 18.1% stress reduction with nine Pareto-optimal designs, validating near-autonomous engineer-level decision making. While promising, the work also highlights practical considerations such as API rate limits, token costs, and privacy, outlining a path toward broader industrial deployment with local LLMs and extended toolchains.

Abstract

In modern engineering practice, human engineers collaborate in specialized teams to design complex products, with each expert completing their respective tasks while communicating and exchanging results and data with one another. While this division of expertise is essential for managing multidisciplinary complexity, it demands substantial development time and cost. Recently, we introduced OpenFOAMGPT (1.0, 2.0), which functions as an autonomous AI engineer for computational fluid dynamics, and turbulence.ai, which can conduct end-to-end research in fluid mechanics draft publications and PhD theses. Building upon these foundations, we present Engineering.ai, a platform for teams of AI engineers in computational design. The framework employs a hierarchical multi-agent architecture where a Chief Engineer coordinates specialized agents consisting of Aerodynamics, Structural, Acoustic, and Optimization Engineers, each powered by LLM with domain-specific knowledge. Agent-agent collaboration is achieved through file-mediated communication for data provenance and reproducibility, while a comprehensive memory system maintains project context, execution history, and retrieval-augmented domain knowledge to ensure reliable decision-making across the workflow. The system integrates FreeCAD, Gmsh, OpenFOAM, CalculiX, and BPM acoustic analysis, enabling parallel multidisciplinary simulations while maintaining computational accuracy. The framework is validated through UAV wing optimization. This work demonstrates that agentic-AI-enabled AI engineers has the potential to perform complex engineering tasks autonomously. Remarkably, the automated workflow achieved a 100% success rate across over 400 parametric configurations, with zero mesh generation failures, solver convergence issues, or manual interventions required, validating that the framework is trustworthy.

Figures (9)

• Figure 1: Engineering.ai
• Figure 2: The four NACA airfoils selected by the Chief Engineer for UAV wing optimization. The design exploration encompasses symmetric profiles (NACA 0012, 0015) for baseline performance and cambered profiles (NACA 2412, 4412) for enhanced lift characteristics, systematically varying thickness (12%-15%) and camber (0%-4%) to explore the aerodynamic-structural-acoustic design space.
• Figure 3: Comprehensive aerodynamic performance analysis across twelve simulation cases. (a)lift coefficient ($C_l$) variation with angle of attack and velocity, with bubble sizes proportional to $C_l$ magnitude. (b)drag coefficient ($C_d$) distribution across operating conditions with bubble sizes representing $C_d$ values. (c)drag polar ($C_d$ vs $C_l$) illustrating aerodynamic efficiency envelopes for each airfoil, with bubble sizes indicating $C_l$ magnitude. (d)pitching moment coefficient ($C_m$) characteristics with bubble sizes proportional to $|C_m|$.
• Figure 4: Acoustic performance comparison showing Overall Sound Pressure Level (OASPL) grouped by operating conditions. (a)OASPL variation with velocity (25-35 m/s), demonstrating identical acoustic levels across all airfoil geometries at each velocity. (b)OASPL distribution across angles of attack ($0^{\circ}$ to $6^{\circ}$), revealing that acoustic emissions are primarily velocity-dependent rather than geometry-dependent.
• Figure 5: Autonomous CAD-to-FEA workflow by the Structural Engineer: (a) Parametric wing geometry in FreeCAD with internal structure. (b) NACA 4412 cross-section. (c) Adaptive tetrahedral mesh by Gmsh with local refinement. (d) Cross-sectional mesh view. This automated pipeline enabled exploration of 432 structural configurations.