4 PhD and 2 post-doc positions at TU Darmstadt on Continual Adaptation

[emtiyaz]

Dear all,

I have open positions available for a new Adaptive Bayesian Intelligence group at TU Darmstadt (Germany) starting Sep. 2026. The deadline for the application is on June 15, 2026.

4 PhD positions: https://www.career.tu-darmstadt.de/tu-darmstadt/job/54256

2 post-doc positions: https://www.career.tu-darmstadt.de/tu-darmstadt/job/54268

Possible research topics are:

• Continual Learning: practical continual learning at scale (for LLMs), building and using memory for continual learning, fast continual learning via adaptation, curriculumn learning, understanding and facilitating plasticity during learning
• Distributed Learning: federated learning at scale, learning to communicate, new algorithms for distributed learning, brain-like learning via local learning rule, multi-modal learning, building a federation of AI model-zoo
• Transparent AI: understanding learning mechanisms of LLMs, knowledge tracing during training, AI transparency via adaptation, concept and representation learning
• Active AI: methods to select datasets and distill models, active model merging and souping, learning to act by quickly adapting, data-efficient reinforcement learning (with application to robotics)
• Approximate Inference: large scale Bayesian inference (for spatio-temporal models and Gaussian Process), PAC Bayes for adaptive intelligence, message passing for deep learning, generalizing Bayesian principles using information geometry.
• Continuous Optimization: fast variants of ADMM for non-convex problems, fast second-order optimization, practical stochastic variance reduction, understanding and generalizing duality principles,
• Applications: NLP, Multimodal optimization, Robotics