项目介绍

2026年7月20日

This research will take place within the lab of Prof. Siegfried Nijssen, which is a research unit of the DTAI section of the Department of Computer Science of KU Leuven. The DTAI section is well-known for its contributions to Machine Learning, Symbolic Artificial Intelligence and combinations of different forms of Artificial Intelligence, including Neuro-Symbolic Artificial Intelligence.

Prof. Nijssen’s lab studies interactions between Machine Learning and Combinatorial Search. Many problems in computer science can be seen as combinatorial search problems. From planning and scheduling problems, to explainable machine learning and computer programming: in all these problems we are looking for symbolic expressions that meet well-defined criteria. However, finding schedules, interpretable predictive models or computer programs that meet well-defined criteria is not easy; these problems are often NP-hard to solve, and many search algorithms have been proposed in the literature to solve these problems. We study how the integration of machine learning and combinatorial search algorithms can help solve such problems more efficiently, where we have a recent interest in how neural networks can be combined with symbolic forms of AI, leading to forms of Neuro-Symbolic AI.

Website van de eenheid

Project

Foundation models have emerged in the last 10 years for many domains: models for text (Large Language Models) and vision models are certainly the most popular at the moment. A more recent development is the emergence of foundational models for more specialized domains, such as Tabular Foundation Models. The core principle of these neural network models is that they are optimized for a specific form of data; in the case of tabular data, for instance, this is data that consists of columns where the order of the columns is not important. Foundation models make few-shot learning possible, in which little data is required to perform a new task well. 
Within this project, the goal is to work towards foundation models for search problems: neural network models that are trained to solve combinatorial search problems when relatively small amounts of data or input are provided at inference time. The envisioned approaches consist of two components: a neural network component and a search component, where the neural network component guides the search component. The neural networks will be trained to guide the search component, where the project will consider increasingly larger classes of search problems.
You are expected to contribute both to theoretical research and to applied research in the context of the Flanders AI program. 

Profile

We expect from applicants:

• an excellent Master degree in Computer Science, Artificial Intelligence or a related discipline
• a background in both the theory and the implementation of neural networks (for instance, transformers, CNNs) and/or search algorithms (for instance, branch & bound search, beam search)
• strong programming skills, where experience in languages such as C++, Rust or Java is an asset
• an ability to carry out independent research
• proficiency in English, both spoken and written

Offer

This is a 4 year position as a PhD researcher. The research will be carried out in the DTAI section at the Department of Computer Science of KU Leuven, at the campus in Leuven.

Interested?

For more information please contact Prof. dr. Siegfried Nijssen, mail: siegfried.nijssen@kuleuven.be.You can apply for this job no later than August 12, 2026 via the online application tool