First Largest Workshop on LLMs Empowering Future Industry at TUM Heilbronn

What does it take to move large language models beyond impressive demonstrations and turn them into technologies that can support real industrial processes? This question shaped two days of talks, discussions, and exchanges at the first workshop on “Large Language Models Empowering the Future of Industry”, held in Heilbronn from 20 to 21 July 2026.

Prof. Dr. Sven Mayer and Ankur Bhatt attended the workshop, which brought together researchers and industry experts to discuss how LLMs and related AI technologies can support industrial processes and decision-making. The event was hosted by Prof. Ziyue Li and Prof. Alena Otto from the Technical University of Munich, together with Prof. Sascha von Behren from TU Dortmund University.

The workshop opened with welcoming remarks from Ziyue Li, Alena Otto, and Sascha von Behren. They introduced the program and highlighted the growing relevance of LLMs in industrial contexts. 

In the opening keynote, Prof. Fugee Tsung presented “Intelligence-Driven Quality Innovation: From Industrial Informatics to Intelligent Decisions.” He discussed how industrial data and intelligent methods can be combined to improve quality management and support better decision-making. The discussion then shifted from potential to implementation. In “Operationalizing LLMs in Industry,” Matthias Baier focused on the practical work involved in bringing LLMs into real industrial environments. Prof. Ran Jin, joining online, continued the manufacturing perspective with the talk “LLMs for Process Modeling in Manufacturing.” A panel discussion titled “Burning Challenges Meet Proven LLM Capabilities” brought speakers from academia and industry into conversation. Rather than looking only at what LLMs might eventually achieve, the panel examined what they can already offer, which industrial challenges remain unresolved, and what gaps must still be addressed before wider adoption becomes possible.

The afternoon broadened the technological perspective further. Prof. Xuan Wang presented “Towards Small, Open-Source, Multimodal LM Agents for Science and Society,” exploring the potential of smaller and more accessible multimodal language-model agents for scientific and societal applications. Dr. Sven Franke from ZIETA followed with “Deep Learning-Enhanced Soft Sensors for Real-Time Glucose Concentration Monitoring in Fed-Batch Bioreactors,” demonstrating how deep learning can support real-time glucose monitoring in bioreactors.

After a day filled with talks and discussions, the poster session created space for more direct exchange. Participants had the opportunity to learn about one another’s work, discuss ideas in greater depth, and make new connections. The first day concluded with a dinner hosted by Prof. Ziyue Li.

The second day began with Prof. Youness Dehbi, who presented “Urban Digital Twins for Smart Cities.” His talk offered insights into how digital twins can help us understand cities and support smarter approaches to urban planning. Prof. Dr. Sven Mayer then gave the talk “Designing Human-Centered Computing Systems in the Age of LLMs.” He addressed the need to think beyond prompts and simple interactions with language models. The talk considered how machine learning and other technologies can be combined to create systems that genuinely support people and respond to their needs. In the final talk, Dr. Mohammadreza Fani Sani discussed how agentic AI can support and automate complex organizational processes. The workshop concluded with a discussion about how the event could be further developed and improved.

Overall, the workshop provided valuable insights for researchers and industry experts to exchange perspectives, discuss open challenges, and explore how LLM-based technologies can be developed for practical use.