HCMC Parternship
Parternship over several projects with the Hellenic Capital Market Commission (HCMC - Επιτροπή Κεφαλαιαγοράς).
The Machine Learning for Humans (MLH) Theme is a research group within the Archimedes/Athena Research Center. Our research focuses on the development and application of Machine Learning (ML) and Natural Language Processing (NLP) methods to address challenges in the humanities, cultural heritage, and related interdisciplinary domains. Our work spans topics ranging from the recognition and analysis of ancient manuscripts and historical texts to biomedical image analysis and other data-intensive applications.
The MLH Group is also actively involved in the organization of the 3rd Greek Olympiad in Artificial Intelligence (PDTN 2027). The Olympiad aims to inspire and engage young talents in Artificial Intelligence (AI), foster computational thinking and problem-solving skills, and encourage students to explore cutting-edge AI concepts and techniques.
Parternship over several projects with the Hellenic Capital Market Commission (HCMC - Επιτροπή Κεφαλαιαγοράς).
Mechanism-Informed Multimodal Generative AI for Causal and Dynamical Modelling in Biomedical Research European Horizon-RIA project (2026-2030).
Exploring LLM-based Mediators in Online Discussions.
Since 2012, the response to toxic and hateful speech has been removing comments and administering disciplinary measures (“content moderation”). In 2026, it seems obvious that this approach is not enough to stem the tide of hateful rhetoric, trolling, and harassment. A proposed alternative suggests using human intermediaries (“facilitators”) who manage the discussion while actively participating, although this solution is not scalable in the modern social media landscape.
Our group is working on automating some of the responsibilities of facilitators by using LLMs. We describe what responsibilities human facilitators typically have, and which can be reasonably automated with LLMs (survey paper), typical failure modes of LLM facilitators through simulated experiments (simulations paper), as well as how LLM facilitation behavior correlates with humans (intervention paper).
Coorganizer of the 3rd Greek Olympiad in Artificial Intelligence (PDTN 2027).
The competition is held annually in two phases: Talos, the qualifying round, and Pandora, the final round, where the most successful participants compete for a place on the Greek National AI Team. The top-performing students are selected to represent Greece at international Artificial Intelligence Olympiads, gaining valuable experience and the opportunity to compete and collaborate with talented peers from around the world.
Using artificial intelligence and LLMs to improve clinical data analysis, patient safety, and healthcare quality in Intensive Care Units (ICUs).
The 2025 project develops an Automated Diagnostic Shift Detection (ADSD) framework to identify changes in diagnoses across consecutive clinical reports, distinguishing between diagnostic stability, refinement, progression, and contradiction. The 2027 project focuses on automated de-identification of Greek ICU clinical reports to protect patient privacy and ensure GDPR compliance. It explores pretrained language models, LLMs, and synthetic data augmentation to overcome the scarcity of annotated Greek clinical data and enable accurate, privacy-preserving, and locally deployable solutions.