The world’s first "AI Engineer" comes from Stuttgart

A research team from the SimTech Cluster of Excellence and the Faculty of Aerospace Engineering and Geodesy at the University of Stuttgart led by Dr. Xu Chu has developed an AI-based engineering assistent capable of autonomously performing simulation and analysis tasks in fluid mechanics. The system, known as "AI Engineer", is believed to be among the first AI-powered engineering assistants capable of autonomously performing complex simulation tasks in fluid mechanics. The "AI Engineer" is based on a multi-agent system and the freely available software package OpenFOAM®, which can be used to simulate flow problems. But the innovation doesn’t stop there. The AI also writes scientific articles independently without human input.

The "AI Engineer" consists of four AI agents that work together and complement each other to solve complex tasks. An AI agent is a software system that can think logically, learn or make decisions, and simultaneously process large amounts of information from text, video, or images. The "AI engineer" combines a large language model, which generates text, with the open-source OpenFOAM® software, which performs numerical fluid mechanics simulations. “Our 'AI engineer' works thoroughly and reliably—like a German engineer,” says Dr. Xu Chu, who has lived in the Baden-Wuerttemberg region for about 20 years and knows the local mentality well. He and his team developed an "AI engineer".

Four agents share the tasks of the "AI Engineer": The preprocessing agent analyzes the user’s requests and forwards them depending on their complexity. The prompt generation agent creates instructions, which it places in a prompt pool. The "AI Engineer" agent interprets the prompts, automatically configures the simulation, and initiates the process. The post-processing agent then analyzes the simulation results and creates comparison diagrams and charts.

To test the reliability of the "AI Engineer", the scientists carried out various experiments. They selected five case studies covering a wide range of computational fluid dynamics tasks. Examples included simulating the flow of a viscous liquid through a pressurized straight channel, modeling multiphase flow in porous media (such as oil drainage), and analyzing turbulent airflow around a motorcycle at different speeds.

Because reproducibility is an indicator of reliability, some simulations were repeated up to one hundred times, and the result was always the same. “That surprised us—and also scared us a little,” says Chu. “For example, if I upload a photo of myself to ChatGPT and ask ten times whether I look good, I get ten different responses. Of course, that’s not possible with engineering issues,” he says. “We’ve lost plenty of sleep realizing the implications of a system that works this reliably,” says Chu, who believes it could revolutionize the field and make intelligence widely accessible.

Xu Chu has taken it a step further by integrating additional agents that read scientific books and publications, develop research ideas, analyze "AI Engineer" results, and write scientific manuscripts—all fully automatically. He calls the system “Turbulence.ai”, and it is the world’s first AI scientist in the field of fluid mechanics. It poses and formulates hypotheses, plans simulations and checks them for novelty and feasibility, corrects errors, prepares the results visually, and finally writes a complete scientific article. The first manuscript on two-phase displacement processes in porous media is already finished. “Because many questions in fluid mechanics remain unanswered, the AI scientist could make a real contribution to the field,” says Xu Chu.

Xu Chu comes from China and has lived in Stuttgart for around 20 years. He completed a bachelor’s degree in automotive engineering at Tongji University in Shanghai. Because he was determined to obtain a degree in engineering from a German university, which stands for high quality and reliability, he studied automotive engineering at the University of Stuttgart. He later earned a doctorate in mechanical and energy engineering at the University of Stuttgart. He conducted research within the DISS project, served as a junior group leader at the Institute of Aerospace Thermodynamics, and was a habilitation candidate at SimTech in the PN1 project network. He has been a Senior Lecturer at the University of Exeter in England since May 2024 and completed his habilitation at the University of Stuttgart in January 2025.