Postdoctoral Fellowship in Predictive World Models for Image Generation Vilnius University in Lithuania

Vilnius University is offering a 12-month postdoctoral fellowship in Predictive World Models for Image Generation within the AI Methods Lab at the Institute of Computer Science, Faculty of Mathematics and Informatics. The project sits in the broad area of computer science and informatics, with a strong emphasis on deep learning, self-supervised learning, generative modelling, and multimodal AI. The successful candidate will work on methods that predict future visual states in latent space and then render them as high-fidelity images, with applications to large-scale image collections such as satellite image time series.

The research agenda includes self-supervised joint-embedding predictive architectures, latent-space forecasting, conditioned diffusion, flow matching, and other techniques for controllable visual generation. The postdoctoral fellow will also investigate how to combine world models with generative backbones, how to use vision-language and structural conditioning signals, and how to improve long-horizon consistency, robustness, and evaluation methodology. The work is methodological, but it is grounded in practical experimentation on large image datasets and implemented in reproducible PyTorch-based pipelines on HPC infrastructure.

Supervision is planned by Assoc. Prof. Dr. Valentas Gružauskas and Assoc. Prof. Dr. Linas Petkevičius. The postdoctoral fellow will join a collaborative and international research environment, with opportunities to contribute to scientific publications, conference presentations, open-source software, and competitive research proposals. The advertisement also highlights mentoring and knowledge-transfer activities, including support for student projects and lab seminars.

The fellowship is full-time, with a gross monthly salary of €3,634. The contract duration is 12 months, and candidates must be able to work on site in Vilnius, Lithuania. Eligible applicants must hold a doctoral degree, preferably from a foreign institution, or from another Lithuanian science institution outside Vilnius University. As a postdoctoral fellowship, it is targeted at researchers within five years of PhD award, with possible extensions for documented parental leave. Strong preparation in deep learning, computer vision, generative modelling, Python, PyTorch, self-supervised learning, and English communication is expected.

Applications must be submitted by email to [email protected] with the subject line "postdoctoral fellowship". The required package includes the official application form, CV, list of up to 10 key publications or patents, and a digital copy of the PhD diploma. The deadline is 7 September 2026.