Postdoctoral positions: Intelligent Robotics / House of AI
About this role
Up to three postdoctoral positions are available in the Intelligent Robotics group at Aalto University, Finland, working on robotics and physical AI in direct collaboration with leading industrial partners. The group develops intelligent robotic systems, with an emphasis on methods and systems that cope with imperfect knowledge and uncertain sensing.
The positions are linked to the AI in Manufacturing and Machine Industry theme of the Aalto House of AI, led by Professor Ville Kyrki. The theme brings together robotics, machine learning, and electrical and mechanical engineering to work on autonomous operation of machines, advanced control, and new data-driven services.
Our industrial partners are global market leaders in their product categories, with installed fleets operating worldwide. The collaboration gives access to production machines and field data, problems defined jointly with their engineers, and pilots run on partner sites.
The group also sits inside a dense AI research environment. Aalto coordinates the Finnish Center for AI (FCAI) where Professor Kyrki leads the Autonomous AI program. The group is also part of ELLIS Institute Finland, the second institute of the pan-European ELLIS network, based on the Aalto campus. Postdocs in the group take part in the seminars and the Europe-wide network the institute brings, alongside compute on LUMI, EuroHPC's flagship supercomputer.
Topics
We are looking for candidates in the following two areas. Funding is in place for all three positions.
1. Robot learning for manipulation: imitation and reinforcement learning of contact-rich, deformable object, and multi-object manipulation, and data-efficient adaptation to new tasks and embodiments.
2. Fault prediction and reliable autonomy: data-driven condition monitoring, prognostics, and anomaly detection for machines and fleets, learning from sparse failure data, handling distribution shift across machines and operating conditions, and combining physical models with