Postdoctoral researcher – latent variable models for creative language generation

About the role

The Centre for Computational Linguistics at KU Leuven seeks a postdoctoral researcher to integrate latent factorization models with neural language models for creative language generation. The project is part of the ERC Consolidator Grant TENACITY, which aims to develop interpretable and creative computational language models.

Key responsibilities

• Develop variational inference methods for transformer‑based language models, leveraging latent structures from tensor factorization models.
• Investigate the interpretability of the resulting latent space and its use for creative generation such as metaphor and style transfer.
• Provide day‑to‑day guidance to PhD students working on related tensor‑factorization and constraint‑based neural generation tasks.
• Publish research findings in top‑tier NLP and machine‑learning venues and present results at international conferences.

Required profile

• PhD in natural language processing, machine learning, computational linguistics, or a closely related field, with peer‑reviewed publications.
• Strong background in deep learning for NLP; experience with latent variable models or variational inference is a plus.
• Excellent programming skills in Python and PyTorch, and experience with large‑scale GPU experimentation.
• Experience supervising junior researchers or a clear ambition to develop such mentorship skills.
• Fluent English with strong oral and written communication abilities.

Required skills

• Python
• PyTorch
• Transformer‑based language models
• Variational inference
• Tensor factorization models
• GPU infrastructure for large‑scale experiments

What we offer

• Full‑time, fixed‑term postdoctoral contract for 2 years.