About Me
I am a Ph.D. student in Digital Innovation and e-Health (started in October 2026) at the University of Bari Aldo Moro, conducting research within the Computational Intelligence Laboratory (CILAB) at the Department of Computer Science.
My research is centered around Artificial Intelligence for Digital Mental Health (DMH). Traditional psychiatric and clinical assessments still rely predominantly on episodic visits and standardized psychometric questionnaires, providing only an incomplete, episodic snapshot of an individual's mental state. Digital Mental Health opens the transformative opportunity for continuous, ecologically valid monitoring via wearable sensors and digital technologies capable of acquiring physiological, behavioral, acoustic, and contextual biomarkers.
While AI offers tremendous promise for analyzing these heterogeneous, high-dimensional streams, its real-world clinical adoption faces major obstacles. My doctoral research is dedicated to addressing these through three core methodological pillars:
Overcoming clinical annotation scarcity through benchmarked comparative analyses of self-supervised learning (SSL), semi-supervised learning, transfer learning, and modern foundation models on physiological time-series (e.g. HRV, EDA).
Investigating the distinct predictive contributions of multiple data modalities (physiological, acoustic, behavioral, contextual) and formulating advanced multimodal fusion architectures capable of handling asynchronous and heterogeneous data flows.
Bridging the black-box gap by comparing post-hoc explainability techniques (LIME, SHAP, counterfactuals) with intrinsically interpretable models (such as fuzzy linguistic representations), validated through human-centered studies involving clinical specialists.
The overarching goal is to formulate a unified methodological framework for designing reliable, robust, and transparent AI systems for continuous mental health monitoring, advancing their translation into clinical practice.
Research Interests & Focus
Publications
Education & Theses
Investigated Self-Supervised Learning (SSL) and modern sequence models (Transformers, State Space Models) for continuous stress detection and mood monitoring using wearable biosignal time-series (HRV, EDA) and Explainable AI.
Designed a non-invasive screening pipeline using Vision Transformers (ViT) on palpebral conjunctiva photography across heterogeneous international cohorts, achieving 82% accuracy in binary anemia classification.