PhD position: Improving Language Models with Formal Knowledge in Industry KU Leuven in Canada

KU Leuven is offering a PhD position in the context of the VLAIO TETRA project Reliable Language Intelligence for Knowledge (ReLIK). The project is based in the DTAI research unit, part of the Department of Computer Science, and the EAVISE research group in Sint-Katelijne-Waver, Belgium.

The PhD project focuses on improving language models with formal knowledge in an industrial context. Modern large language models are powerful for building intelligent assistants and company-specific search functions, but they still raise concerns about reliability because they can hallucinate facts and do not guarantee correctness. In contrast, knowledge graphs and formal reasoning engines are reliable at reasoning but less flexible with ambiguity. The research goal is to combine these approaches into neuro-symbolic AI systems that are more reliable and better suited for industry.

The exact direction of the PhD will be shaped by the needs of the industrial consortium and the candidate’s own interests. The work will involve examining company questions and requirements, identifying open problems, and developing new methods and techniques to remove barriers to industrial deployment. This makes the position especially relevant for candidates interested in language models, knowledge representation, formal reasoning, and applied artificial intelligence.

The successful candidate will join the EAVISE group at KU Leuven in Sint-Katelijne-Waver, a location described as centrally located in Flanders and accessible by train or car. KU Leuven offers a driven research environment, a dynamic research group, access to computer infrastructure and travel budgets, and opportunities to attend international conferences and build a scientific network.

Applicants should have a master’s degree in computer science, AI, informatics, or a related field, along with solid experience using language models in applications. Familiarity with formal systems is an advantage but not required. Fluent English is required, and Dutch is a helpful bonus. The ideal candidate is motivated by neuro-symbolic AI and the goal of making AI systems more reliable, and should collaborate well with others.

The position is advertised as a full-time PhD post for one year, with the possibility of extension up to four years after a positive evaluation. The application deadline is 2026-08-24 23:59 (Europe/Brussels). Interested candidates can contact prof. dr. Simon Vandevelde at [email protected] and apply through the KU Leuven job page.