There will be one PhD opportunity for a highly motivated PhD student to develop a human-centered world model for intelligent systems that interact with people. This will be one position in my new group at the ELLIS Institute Finland and the University of Turku.
The project will investigate how humans, their behaviors, and their interactions with the environment can be represented and modeled within generative world models. A central research direction is to build world models in which humans are active participants rather than passive elements of a scene, enabling the modeling and generation of realistic human behaviors, interactions, and responses in dynamic environments.
The ultimate goal is to use these models as scalable data-generation and simulation platforms for human-robot interaction, allowing intelligent systems to learn from diverse, controllable, and realistic human interaction scenarios beyond what can be practically collected in the real world.
Depending on the candidate's interests, the research may include generative world models, multimodal foundation models, human behavior modeling, video generation, simulation, and human–robot interaction.
Research Areas
Human-Centered World Models
Generative Models and Multimodal Foundation Models
Human Behavior and Interaction Modeling
Data Generation and Simulation for Human–Robot Interaction
Embodied AI and Human-AI Interaction
This is a four-year fully funded PhD position at the ELLIS Institute Finland & the University of Turku, and you will work in a highly collaborative research environment. We will work closely with ETH Zürich and leading industrial research institutes through the project development.
We welcome applicants with a strong background in computer vision, machine learning, generative AI, robotics, or related fields.
Starting date
As soon as possible, from October 2026.
Application
Please only send applications to this email: xucong.zhang@utu.fi
Please do NOT send your application to my other email address.
Please submit your CV, academic transcripts, a one-page research statement, and the names of two referees.