Our work around eDoer focuses on developing the next generation of personalized online learning systems. The research responds to the growing demand for high-quality, adaptive, and skill-oriented digital education. Today's learners expect more than just content. They seek platforms that help them acquire job-relevant skills, explore personal interests, and access curated knowledge across diverse domains. To meet these needs, our research investigates several key questions and themes:
Our multidisciplinary approach integrates research in technology-enhanced learning, artificial intelligence, user modeling, and software engineering to create a dynamic, engaging learning ecosystem through eDoer.
A compact overview of selected work on AI-supported learning, open educational resources, curriculum development, and personalization.
We introduced eDoer as an AI-driven recommender that connects learner goals, labor-market skills, and open educational resources, and validated it in a randomized controlled trial (RCT) with learners.
Healthcare relevance: The same logic can support role-based health staff onboarding and continuing education.
We benchmarked embedding-based rankings for learning-outcome alignment, validated them with experts, and linked them to learner performance in a three-group randomized controlled trial (RCT).
Healthcare relevance: For healthcare, this helps check whether a resource really supports a required competency or patient-learning goal.
We developed a human-AI workflow for personalized curriculum creation and validated the platform in two pre-post randomized controlled trials (RCTs) comparing it with Moodle on knowledge acquisition.
Healthcare relevance: This is highly relevant for healthcare, where experts need trustworthy AI support to build and validate learning paths quickly.
We designed and evaluated LLM-assisted interfaces that reduce prompt-engineering effort for curriculum authors.
Healthcare relevance: This supports busy educators and health professionals who need practical authoring tools.
We studied how LLMs can generate learner-specific answers and how those answers can be evaluated automatically and by humans.
Healthcare relevance: This points toward clearer explanations for patients, relatives, staff, and other learner groups.
We evaluated whether LLMs can generate useful course topics and help refresh curricula beyond conventional search workflows.
Healthcare relevance: The approach can help experts draft first versions of training paths in healthcare and other domains.
We analyzed OER metadata and showed how metadata signals can help predict resource quality across repositories.
Healthcare relevance: This matters when patient or staff education depends on trusted, well-described learning content.
We built a prototype that links job-skill requirements with open educational videos and recommends relevant learning content.
Healthcare relevance: This can transfer to clinical training videos, nursing skills, and practical onboarding paths.
We combined AI recommendations and expert input to help contributors create and update personalized curricula.
Healthcare relevance: This fits fast-changing healthcare knowledge, where experts must keep learning paths current.
We defined and evaluated quality metrics that help authors and platforms assess open educational resources.
Healthcare relevance: Quality checks are especially valuable for sensitive learning contexts such as patient education.
We contributed to an ontology for representing learning resources, skills, user profiles, and recommendations in a connected structure.
Healthcare relevance: A shared structure makes it easier to model complex domains such as healthcare without losing cross-domain flexibility.