You will play a central role in the development of Large Language Model (LLM) components that support education professionals in creating, maintaining, and improving Individual Development Perspective Plans (OPPs). The project does not aim to automate the creation of complete OPPs. Instead, you will investigate how LLMs can best support education professionals with specific, well-defined subtasks within the OPP workflow.
Examples include generating a first draft of the Integrative Student Profile, supporting the formulation of the Educational Perspective, and providing targeted feedback on the quality of existing OPP text. A key research challenge is to determine how information from different OPP sections can be represented and provided to an LLM so that the generated output is coherent, concrete, actionable, and consistent with the underlying information.
You will develop and evaluate prompting strategies, structured input representations, and, where useful, fine-tuned models. You will also work on LLM-based quality feedback, where a quality rubric defines criteria such as concreteness, actionability, coherence, and specificity of support needs to identify weaknesses in OPP text and provide concise, prioritized, and practically useful suggestions for improvement.
An equally important part of the position concerns the interaction between the AI component and the education professional, where the AI system should support professional judgment rather than replace it. Together with a software development partner, you will help design how these AI components can be integrated into an existing student information system. The research will be conducted iteratively and in close collaboration with education professionals.
What you do
- Select and benchmark state-of-the-art open-source and commercial LLMs for Dutch-language OPP-related tasks
- Develop and evaluate LLM components for generating specific OPP sections, initially focusing on the Integrative Student Profile and the Educational Perspective / justification of educational pathway
- Develop prompting strategies and structured input representations that combine information from relevant OPP sections
- Investigate when prompt engineering is sufficient and when techniques such as fine-tuning or retrieval-augmented generation provide added value
- Develop rubric-based LLM methods for detecting quality issues in existing OPP text and generating concise, prioritized improvement suggestions
- Develop methods for identifying generic or baseline support that occurs repeatedly across OPPs and distinguish this from student-specific support needs
- Design and evaluate human-AI interaction patterns for AI-assisted OPP writing and revision
- Investigate how generated text and feedback can best be presented so that education professionals retain control over the final content
- Collaborate with software developers who will develop the technical architecture and API-based integration of LLM components into a student information system
- Define which OPP data elements should be transferred to an LLM component and how generated output should be returned and incorporated into the existing workflow
- Evaluate LLM output using the OPP quality rubric developed in the project and through evaluation with education professionals
- Conduct co-design and user studies with teachers, behavioural scientists, support coordinators, and other education professionals
- Analyze the quality, reliability, and limitations of generated OPP text and AI feedback
- Contribute to open-source implementations of reusable OPP-specific LLM components and evaluation tools
- Publish results in peer-reviewed AI, NLP, Human-AI Interaction, and/or Educational Technology venues
- Supervise MSc students working on related research topics
- Collaborate closely with project partners from education, research, and software development
What they ask
- PhD in Artificial Intelligence, Computer Science, Computational Linguistics, Natural Language Processing, Human-AI Interaction, Data Science, or a closely related field
- Ability to work in a multidisciplinary consortium involving researchers, software developers, and education professionals
- Excellent communication skills in English; proficiency in Dutch, or the willingness and ability to develop sufficient Dutch proficiency to work with Dutch-language educational materials and users, is strongly preferred
- Affinity with education, special education, or educational technology is highly desirable
- Demonstrated hands-on experience with Large Language Models and modern NLP methods
- Good programming skills, preferably in Python, and experience with relevant ML/NLP frameworks such as PyTorch and Hugging Face Transformers
- Experience with prompt engineering and systematic evaluation of generative AI systems
- Ability to design and run empirical comparisons between different models, prompting strategies, and system configurations
- Experience with qualitative and/or quantitative evaluation of generated text
- Interest in Human-AI Interaction and in designing AI systems that support rather than replace professional decision-making
- Experience with Retrieval-Augmented Generation, model fine-tuning, structured generation, LLM evaluation frameworks, or human-centered AI is an advantage
- Experience with co-design, user studies, or iterative prototyping with domain professionals is an advantage
- Analytical and scientific writing skills
Practical
- Salary between € 3,706.00 and € 5,760.00 gross per month on a full-time basis (Scale 10)
- A full-time 38-hour working week
- 232 hours of holiday leave per year for full-time employment
- 8% holiday allowance and 8.3% end-of-year bonus
- Solid pension scheme (ABP)
- Contribution to commuting expenses
- Possibility of hybrid working
How to apply
View the full assignment text and application details once your tailored application is ready.
Order a tailored application to view the full assignment and application details.
More context, less searching.
You get enough context to judge whether this job is relevant. The full brief, client details and next steps stay available inside the app.