📌 Key facts
- What: This Master’s thesis or IDP builds on our published research on regulatory change detection. Using an existing system and benchmark, you develop and evaluate AI methods that determine which changes in EU regulations affect a specific organization and what they imply.
- When: Start anytime soon!. Applications are open!
- How to apply: Send your CV, transcript of records and a motivation letter, with the subject [IDP] RegRadar or [MA] RegRadar (more details below).
💡 Background
Organizations must decide for every amendment to the regulations they are subject to whether it requires action. Doing this manually is time-consuming, error-prone and depends on scarce legal expertise, which makes it a prime use case for applied AI. In prior work, we built RegTrack, a benchmark for fine-grained change detection in EU regulations with 4,772 annotated provision pairs and six change classes (ACL 2026). We also built an end-to-end system that automatically aligns, classifies and explains changes between two versions of a regulation, combining embeddings, staged LLM classification and a graph database (ICIS 2026).
We then tested the system with compliance professionals, and one finding stood out. The explanations were good enough to confirm that something changed, but not to decide what it means for their own organization. Reviewers wanted to know things like: Does this change affect our deadlines, our obligations or our reporting duties? To answer that, a system needs to know the organization, and current approaches don’t take that into account. This project closes that gap.
🦾Who We Are
I am a PhD candidate in the BMW doctoral program and a Research Associate at the Chair for Strategy and Organization at TUM. My research explores the application of generative AI across diverse business domains, with a particular focus on the design, development and evaluation of LLM systems. I am especially interested in data quality, system reliability, and the performance of LLM agents in retrieval and extraction contexts. I have an educational background in Information Systems (B.Sc., M.Sc., TUM) and have published and accepted papers on regulatory change detection and LLM-based document analysis at venues such as ACL, ICIS, HICSS and PACIS.
The Chair for Strategy and Organization is focused on research with impact. This means we do not want to repeat old ideas and base our research solely on the research people did 10 years ago. Instead, we currently research topics that will shape the future. Topics such as Agile Organisations and Digital Disruption, Blockchain Technology, Creativity and Innovation, Digital Transformation and Business Model Innovation, Diversity, Education: Education Technology and Performance Management, HRTech, Leadership, and Teams.. We are always early in noticing trends, technologies, strategies, and organisations that shape the future, which has its ups and downs.
🎯 Goals
You will extend the existing system with the step from a detected change to its consequence for a concrete organization:
- Model organization profiles, such as a vehicle manufacturer, a supplier or a shipping company, derived from the actor categories the regulations themselves define.
- Build an information extraction pipeline that turns each provision into a structured representation of the obligation (who must do what, under which condition, by when) and compares two versions field by field. You are free to choose and combine methods, from LLMs with structured outputs to fine-tuned encoder models, embedding-based approaches or rule-based extraction.
- Design and benchmark relevance approaches, for example generative models given a free-text company profile against a structured matching of profile and provision scope, or classical ML and NLP baselines.
- Set up the data and annotation workflow to build a high-quality evaluation set together with domain experts.
- Run a rigorous evaluation with precision, recall and F1, and analyze where and why models fail.
Guiding research questions:
- How accurately can automated methods identify which regulatory changes affect a specific organization, compared with expert judgment?
- Does a structured representation of obligations lead to more accurate and more verifiable implications than free-text answers from generative models?
🎓 Profile
- Solid Python skills and experience with data processing (e.g., JSON/XML, pandas)
- First experience with or strong interest in NLP and machine learning (e.g. information extraction)
- Ability to work independently and in a structured way
Nice to have
- Experience in evaluating ML/NLP systems or designing experiments
- Experience in data annotation or building labeled datasets
- Familiarity with databases (SQL, vector or graph databases) or web development
- Interest in Legal Tech, regulation and compliance (no legal background required)
- Curiosity for interdisciplinary problems between technology, law and organizations
📚 Further Reading
- Yu, J., Li, K. C., & Ostarek, J. (2026, July). RegTrack: A Fine-Grained Benchmark for Multi-Class Legal Change Detection. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) (pp. 764–778). https://doi.org/10.18653/v1/2026.acl-srw.68
- Abualhaija, S., Ceci, M., Sannier, N., Bianculli, D., Briand, L. C., Zetzsche, D., & Bodellini, M. (2024). AI-enabled regulatory change analysis of legal requirements. In 2024 IEEE 32nd International Requirements Engineering Conference (RE) (pp. 5–17). https://doi.org/10.1109/RE59067.2024.00012
📝 How to Apply
Send an email to joe.yu@tum.de with:
- your CV
- your current grade report
- a motivation letter (max. one page) describing why the topic interests you, your relevant background (e.g., in NLP, machine learning or data processing) and any prior exposure to Legal Tech, regulation or compliance. No legal background is required, but tell us what draws you to the intersection of AI and law.
Please use the following email subject, depending on whether you’re applying for an IDP or a Master’s thesis:
- [IDP] RegRadar
- [MA] RegRadar
We're greatly looking forward to hearing more about you!
📬 Contact
Joe Yu, Chair for Strategy and Organization, Technical University of Munich