to be filled from
03/01/2026
Your working environment
At MAN Truck & Bus, we are part of a strong international team and one of Europe's leading commercial vehicle manufacturers and providers of transportation solutions. Together with Scania, Volkswagen Truck & Bus, International we are part of the TRATON GROUP.
As part of this group, we face major challenges: Our vehicles are becoming increasingly autonomous and connected in order to reduce emissions on the road. We are working on sustainable solutions for this. A career at MAN Truck & Bus offers countless opportunities to participate in this change. Putting the customer first, respect, team spirit, responsibility and avoiding waste - these are the corporate values we live by at MAN.
Pull together with us. As part of our global team of over 36,000 employees, you will join us in turning visions of the future into reality.
We see the individuality of our employees as a strength and welcome diverse applications from people with different backgrounds. If you need support during your application, please contact us.
You can look forward to these tasks
Depending on the current project phase, you will support us in a range of exciting areas that can be tailored to your interests and skills. Your day-to-day work will involve contributing to an ongoing prototype of a compound LLM System; the thesis itself will focus on analysing how compound multi-agent LLM systems can optimally transition from proof-of-concepts to production-ready solutions. In this context, your responsibilities may include:
• Collaborating on prototype features of a multi-agent LLM engineering assistant to deepen understanding of architecture and workflows
• Supporting evaluations, benchmarking and structured testing
• Participating in internal workshops and contributing to system demonstrations
• Analysing the current prototype architecture and identifying gaps toward production readiness
• Deriving recommendations for improving robustness, maintainability and scalability of Compound LLM systems as a whole
• Researching state-of-the-art academic and industrial approaches in order to identify success factors for GenAI automation projects in industry
• Working with key technologies including Amazon AWS Bedrock, Ansys DPF, custom simulation analysis libraries and embedding models
You will work at the intersection of AI and Engineering, gaining hands-on experience with engineering workflows and modern generative AI systems. Your thesis will centre on understanding and evaluating how multi-agent LLM systems can evolve from proof of concepts into a reliable, industrial-grade system combining your practical project experience with analytical and research-driven insights.
Thesis Topic Background
Compound LLM architectures extend traditional LLM setups by integrating multiple specialised agents, external tools, domain-specific libraries and orchestration logic. These systems enable multi-step, tool-driven workflows that more closely resemble real engineering processes. While promising prototypes exist, the transition from proof-of-concept systems to robust, industrial-grade solutions remains challenging.
These challenges span both technical factors such as reliability, reproducibility, observability, scalability and integration with existing engineering toolchains as well as socio-technical factors, including user acceptance, skill readiness, changes to established workflows, shifts in decision authority and broader questions of trust, responsibility and transparency. Successful adoption requires an understanding not just of system behaviour, but also of how engineers collaborate with AI-driven agents and how organisations prepare for such transformations.
As part of your thesis, you will investigate how multi-agent LLM systems can be designed, evaluated and evolved to meet production-readiness requirements in an industrial engineering environment. The work combines hands-on analysis of an existing prototype with research-driven exploration of architectural patterns, industry practices and academic insights. The goal of this thesis is to systematically investigate what is required to bring compound multi-agent LLM systems from PoC-level prototypes to production-ready workflow automation. In line with this goal, the thesis will address two key areas:
(1) Production-readiness requirements: Derive a structured set of technical guidelines for deploying multi-agent LLM systems in industrial engineering environments.
(2) Human-AI interaction and workflow transformation analysis: Investigate how engineers interact with Compound LLM Systems, how decision-making is augmented or shifted, and which tasks are transformed or displaced. Analyse implications for trust, transparency, cognitive load and responsibility to inform decisions about integrating these systems in organizations.
We are looking forward to
• Ongoing studies in Computer Science, Engineering, Mechatronics, Artificial Intelligence or a related technical field
• Programming experience and motivation to work in applied AI contexts
• Interest in modern GenAI orchestration frameworks (e.g., LangChain, LangGraph)
• Strong ability to work independently, take initiative, and structure your tasks
• Curiosity about LLM-based agents, workflow automation and engineering simulation processes
• English and German skills (both are acceptable, bilingual communication possible)
• Optional but beneficial: familiarity with cloud environments such as AWS
Work modalities
As our team is distributed, remote work is encouraged and fully supported. If you prefer working on-site, office space in Munich is available. Two short visits to Munich for hardware handover (start and end of the thesis) are required.
Job information
Reference number: 5224
Claudia Wünsche, claudia.wuensche@man.eu, will be happy to answer any questions about HR you may have.
The contact person for the department is Maximilian Kretzschmar, maximilian.kretzschmar@man.eu .
Integrity and compliance are essential parts of our corporate culture.
We promote diversity and equal opportunities and look forward to receiving diverse applications.