KTH Royal Institute of Technology’s School of Electrical Engineering and Computer Science is seeking a postdoc in probabilistic machine learning.
Project description
A central challenge in machine learning is ensuring reliability and trustworthiness. This is especially crucial in high-stakes applications such as medical diagnosis, autonomous driving, and agentic systems. To improve the reliability and trustworthiness of large-scale machine learning models (e.g., LLMs) in a meaningful way, we, therefore, need new scalable methodologies that can efficiently and accurately capture, represent, and reason about uncertainties based on rigorous and principled frameworks.
The goal of this position is to conduct independent research and develop theories and methodologies that help in improving the reliability and trustworthiness of large-scale machine learning models (e.g., LLMs) in a meaningful way. You will be joining the research group of Martin Trapp and develop your own research agenda in the context of the group’s research. A particular focus of the group is to develop and utilise theories from tractable models (probabilistic circuits) and Bayesian statistics to tackle the reliability of machine learning models, touching topics such as uncertainty quantification in large-scale models, exact and approximate inference, and neurosymbolics. See Martin Trapp’s website for details. The candidate is expected to augment the research expertise of the group through her/his own experience and to conduct research in close collaboration with the group. You will develop theoretical and methodological work that will be released as open-source libraries and publish your work at top-tier machine learning venues (NeurIPS, ICML, ICLR, UAI, AISTATS). Moreover, you will have the opportunity to collaborate with researchers working on machine learning at KTH and collaborate with international experts through the group’s network.
