Ловите несколько позиций
Want to work with me? And be part of a small but dedicated, well-connected (and genuinely awesome 🤓) team that published 6 EMNLP papers and 4 ACL papers this year alone? Then now's your time, because together with colleagues across DFKI, we're recruiting three positions across three complementary research areas!
1. PhD Researcher: Mechanistic Interpretability for Safe Agentic AI
Development of fine-grained steering methods to control language model behavior without retraining, advancing post-hoc controllability through circuit analysis and sparse feature steering.
https://lnkd.in/eS8QnKvJ
2. PostDoc: Multilingual Mechanistic Interpretability
Systematic study of reasoning circuits in compact multilingual models, investigating circuit transfer across scales and 20+ languages. Direct collaboration with Inria's ALMAnaCH team (Djamé Seddah), with research visits and joint workshops.
https://lnkd.in/e69eF9f2
3. PostDoc: AI Safety, Ethics, and Agentic Systems
Develop safe autonomous agents with explicit, inspectable normative constraints. We're looking for safety, or safety/ethics backgrounds open to interpretablity methods. Tailored supervision across NMM, SAINT, and MLT.
https://lnkd.in/ecYwYpgw
These roles sit within two initiatives: SLIMS, a planned project on circuit structure in multilingual models, and SAgA, on safe agentic AI through structurally preserved normative reasoning. The two projects collaborate in one interdisciplinary team.
The PhD and SLIMS postdoc are supervised by me within MLT (to be directed by Marius Mosbach), with support from SAINT (Dominik Hintersdorf, Kristian Kersting) and NMM (Kevin Baum, Verena Wolf). For the safety/agentic systems postdoc, supervision is flexible across three groups: candidates focused on interpretability work with me, those on formal ethics with Kevin, and those on agentic system design with Dominik.
Applications are open until Sep 15th.
https://jobs.dfki.de/en/vacancy/senior-researcher-postdoc-m-w-d-x-ai-safety-ethics-and-agentic-systems-618587.html
September 1, 2026 1.1K 9
