Several Postdoctoral researcher and 1-2 doctoral researcher positions in machine learning in Kaski Lab, ELLIS Institute Finland and Manchester Centre for AI Fundamentals

Benefits

Additional Information

Samuel Kaski's two-part research lab in ELLIS Institute Finland ( Probabilistic Machine Learning , Aalto University) and the Centre for AI Fundamentals in University of Manchester, is searching for postdocs and doctoral students to work on AI fundamentals in exciting projects. The work includes collaboration within ELLIS Institute Finland , the Finnish Center for Artificial Intelligence (FCAI), with the rest of ELLIS , and researchers from other fields. Samuel Kaski is Professor of Computer Science in Aalto University and Professor of AI in the University of Manchester. He is the Director of ELLIS Institute Finland and the Finnish Center for Artificial Intelligence . His research group develops machine learning principles and methods focusing on a few key topics, often working with researchers of other fields in new exciting applications (see currently available topics below). Topics You will join a team developing the next generation of probabilistic and collaborative AI. We study fundamental questions in machine learning, including uncertainty-aware and simulation-based inference, generative modeling, robustness under distribution shift, automatic experimental design, privacy-preserving learning, (inverse) reinforcement learning, computational rationality, and user modelling. Our goal is to develop principled AI methods that are reliable, adaptive, and scientifically useful. The research combines advances in ML foundations with real-world applications in domains such as scientific discovery, healthcare, and design or drugs, materials, systems. By bringing together expertise in machine learning, statistics, optimization, we tackle challenging interdisciplinary problems that cannot be solved by any single approach alone. Below, we outline the research topics for which we are currently seeking candidates. Multimodal foundation models Key words: multimodal learning, foundation models, human-aligned fine-tuning, fine-tuning for downstream tasks, test-time adaptation You will join a research team developing next-generation multimodal foundation models that can reason across text, images, video, and 3D molecular design and robotic environments . The research is conducted within an EU-funded European AI initiative ELLIOT . Our goal is to make these systems more grounded, adaptable, efficient, and aligned with human goals and feedback. The work combines fundamental advances in multimodal representation learning with practical questions of deploying large-scale AI systems in dynamic real-world settings. Depending on your interests, you may work on topics such as large-scale multimodal training, test-time adaptation under distribution shift, efficient model distillation and adaptation, retrieval-augmented learning, or alignment through human feedback, preference, and interaction. Out-of-Distribution Deployable Machine Learning Key words: out-of-distribution generalization, distribution shift, active learning, human-in-the-loop learning, probabilistic modelling, sequential experimental design, collaborative AI, decision support We develop machine learning methods that remain reliable when deployed outside their training conditions. A central challenge in modern AI is that real-world environments differ from the data that models were trained on, leading to failures caused by distribution shifts, hidden confounders, and incorrect assumptions. Our ERC AdG-funded research addresses these challenges by combining probabilistic machine learning, adaptive inference, and human-collaborative AI. Your work will focus on developing algorithms and frameworks that enable models to adapt to new environments, learn efficiently from limited feedback, and support human decision-making under uncertainty. Depending on your interests, the research may involve out-of-distribution generalization, domain adaptation, active learning, learning from expert feedback, sequential experimental design, collaborative AI systems, or probabilistic approaches to robust deployment. The project combines foundational ML research with opportunities to collaborate closely with leading application-domain experts and international research partners. Collaborative AI Key words : collaborative AI, human-AI interaction, decision support, human-in-the-loop learning, uncertainty-aware AI, interactive machine learning, computational rationality, AI-assisted discovery You will join a research team developing collaborative AI systems that work effectively with people in complex decision-making and problem-solving tasks. Our goal is to build AI methods that can interact naturally with users, reason under uncertainty, adapt to human preferences and expertise, and support reliable human decision-making. The research combines machine learning, probabilistic modelling, cognitive modelling, and interactive AI to develop systems that complement rather than replace human intelligence. You may work on topics such as human-in-the-loop learning, uncertainty-aware dec