Agentic AI & GraphRAG for Model-Based Systems Engineering

Model-Based Systems Engineering (MBSE) is the industry standard for designing complex, interconnected systems—from autonomous vehicles to spacecraft. While MBSE provides a single source of truth and enables consistency checking and model reuse; however, the integration with unstructured data and incorporating world knowledge remain open challenges. Furthermore, user interaction creates a need for intelligent semantic search and proactive recommendations.

Large Language Models (LLMs) and Agentic workflows offer a promising solution to bridge this gap. However, context limits and context rot present challenges that simple RAG approaches fail to overcome.

Recent developments in context engineering (e.g., Anthropic’s agentic search) and GraphRAG variants (e.g., LazyGraphRAG) have significantly advanced LLM capabilities to deal with user-provided data. Yet, how to effectively adapt these techniques to the structured models available in Systems Engineering remains an open question.

We offer thesis topics focusing on turning the formal syntax of Systems Engineering (e.g., SysMLv2) into structure-aware context representations for LLMs. We are looking for students to develop novel architectures where agentic workflows leverage semantic understanding to query model repositories and utilize tool-calling capabilities to perform valid modifications to the underlying engineering models.

Your Profile

  • Current Studies: Master’s student in Computer Science, or a closely related field.
  • Programming: Strong programming skills are essential (i.e., proficient in Python).
  • AI Fundamentals: Prior knowledge of LLMs is required; concepts like embedding vectors and cosine similarity should be well-understood, and you should be proficient in working with them
  • Mindset: Motivated, proactive, and independent. You are ready to contribute your own ideas, not just implement predefined tasks.
  • Communication: Good communication skills in German or English.
  • Readiness to Learn: Ability to quickly grasp new technologies, specifically:
    • Agentic LLM frameworks (e.g., PydanticAI, LangGraph, or CrewAI)
    • Graph Databases (e.g., Neo4j)
    • Working with Abstract Syntax Trees (AST) or SysMLv2 API interactions

What We Offer

  • Freedom & Resources: You have a preferred agentic framework? Or Graph database? Furthermore, we provide access to commercial LLM APIs.
  • Cutting-Edge: Work at the forefront of research designing context engineering strategies for Systems Engineering that stay relevant even as LLM capabilities improve.
  • Publication: We explicitly aim for high-quality results suitable for academic publication.
  • Real-World Impact in cooperation projects with large industrial users of System Engineering / SysML.

How to Apply

For more information or to apply, please send your application documents (CV, Transcript of Records, and a short motivation statement) to: Henrik Thillmann 📧 thillmann@se-rwth.de

We define the exact scope together with you, if you already have preferences, you can say so in your application documents.