The Centre for Machine Learning within the Data Science and Statistics Section of the Department of Mathematics and Computer Science (IMADA) at the University of Southern Denmark invites applications for postdoctoral research fellowship position(s) within the field of machine learning, natural language processing, and AI safety. The proposed starting date is 1 February 2026 or soon thereafter. The appointment will be made for an initial period of 12-24 months, with the possibility of extension, though no longer than a total of 4 years in Denmark, at an internationally competitive salary.

The research will be conducted within the MIST project (Scalable Mechanistic Interpretability for Safe and Trustworthy LLM Agents), recently funded by the Novo Nordisk Foundation. The project aims to develop new scalable methods for understanding the inner workings of large language models and developing functionally-grounded steering and control techniques.

The successful candidate will contribute to frontier research on one or more of the following topics:

• Interpretability and transparency: Developing methods to understand how language models process information and make decisions, such as sparse autoencoders, circuit discovery, activation patching, and representation engineering, with a focus on compositional structure in learned representations, as well as testing universality across models and languages.

• Agentic and multi-agent safety: Understanding and ensuring safe behavior in LLM agents that can plan, reason, and use tools, as well as studying the dynamics, communication patterns, and safety properties of multi-agent systems

• Control and containment: Developing steering techniques and safety measures to guide model behavior, including red-teaming and methods for intervention, guardrails, and safety certificates based on mechanistic understanding. Applications include high-risk domains such as healthcare, where synthetic data generation may be leveraged for safety evaluation.

The ideal candidate has a research background in or research experience with one or more of the following topics:

• Natural language processing & language modeling

• Machine learning & representation learning

• Interpretability and analysis of models

• Alignment and language model agents

• Other backgrounds that could inform language model interpretability and control, such as cognitive science, neuroscience, causal inference, probabilistic graphical models, physics/dynamical systems.

Qualifications:
The candidate is expected to hold (or be about to complete) a relevant PhD degree in Computer Science, Artificial Intelligence or another field that provides a strong research background for the project.

Fluency in English and Python are required.

Research experience working with large-scale machine learning projects, extensive research software development experience, and intimate knowledge of machine learning frameworks (such as PyTorch and Transformers) are advantageous. Publications in top ML/NLP venues such as NeurIPS, ICLR, ICML, ACL, EMNLP are expected.

PhD candidates about to complete will also be considered and should attach a statement from their supervisors regarding their impending completion.
The successful candidate will have the opportunity to contribute to establishing a new research group on AI Safety at SDU and will participate in publishing high-quality research papers at top-tier machine learning and NLP venues such as NeurIPS, ICLR, ACL, and EMNLP.

IMADA has the unique feature of bringing mathematicians and computer scientists together within a single department to foster theoretically well-backed high-quality data science research. IMADA is home to many ongoing externally funded research projects, as well as to a rich curriculum of data science and artificial intelligence courses. The Data Science and Statistics Group is a synergy platform for the data science experts in IMADA.

Place of work: The Department of Mathematics and Computer Science is located at the main campus of the University of Southern Denmark, Odense, Denmark. The University of Southern Denmark was founded in 1966 and now has more than 27,000 students, almost 20% of whom are from abroad. It has more than 3,800 employees, and 115 different study programmes in the fields of the humanities, social sciences, natural sciences, health sciences, and engineering. Its main campus is located in Odense, the third largest city in Denmark.

Odense provides family-friendly living conditions with the perfect combination of a historic city centre with an urban feel and yet a close proximity to beaches and recreational areas. Its location on the beautiful island of Funen is ideal with easy access by train or highway to the bigger cities of Aarhus and Copenhagen. As the birthplace of Hans Christian Andersen, Denmark's famous fairytale author, the city is home to a vibrant and creative population that hosts numerous festivals and markets throughout the year.

Application deadline: 19 December at 23:59 hours local Danish time

Please see the full call, including how to apply, on https://fa-eosd-saasfaprod1.fa.ocs.oraclecloud.com/hcmUI/CandidateExperience/da/sites/CX_1001/job/3412/?utm_medium=jobshare