I am a Marie Skłodowska-Curie PhD student at the University of Cambridge under the supervision of Prof. Cecilia Mascolo. My research focuses on vocal biomarkers for uncertainty-aware grounded conversational AI for health assessment.
Driven by curiosity and a commitment to innovation, I aim to harness AI to address practical challenges and improve quality of life though digital healthcare.
My path, from a BSc and MSc in Computer Engineering (AI/ML) at the University of Calabria to international research and industry experiences in Luxembourg and the United Kingdom, has shaped a multidisciplinary and collaborative approach to scientific inquiry.
Beyond research, I am passionate about sports and outdoor activities.
Research Assistant & PhD Student · University of Cambridge
My work focuses on trustworthy, adaptive, and resource-efficient machine learning for multimodal health monitoring.
Robust respiratory audio QA under real-world sensing variability · MobiUK 2026 · abstract
Investigated respiratory audio question answering for health assessment under the heterogeneity introduced by real-world mobile sensing conditions, with a focus on developing robust systems across diverse recording environments and devices.
Evaluation of heterogeneous respiratory audio biomarkers · IVBC 2026 · abstract
Benchmarked heterogeneous audio biomarkers for respiratory health question answering, examining how complementary respiratory sounds can support reliable and generalizable health assessment.
Hierarchically routed multimodal QA model · arXiv
Proposed RAMoEA-QA, a two-stage specialized generative system for respiratory audio QA: an Audio Mixture-of-Experts routes each recording to the most suitable pre-trained audio encoder, while a Language Mixture-of-Adapters selects a LoRA adapter on a shared frozen LLM to match query intent and answer format, improving robustness under domain, modality, and task shifts.
Respiratory audio question answering benchmark · arXiv
Curated and harmonized 11 respiratory audio corpora to build RA-QA, the first respiratory audio question-answering dataset—about 9M QA pairs across 60+ attributes and three formats (verification, multiple-choice, open-ended)—and introduced a benchmark comparing audio-text generative models against traditional audio classifiers.
Lent Term (2026/27) · 2 students from Corpus Christi College
Michaelmas Term (2026/27) · 6 students from King's College
Lent Term (2026/27) · 8 students from Corpus Christi, Clare, Selwyn and Emmanuel Colleges · Syllabus
Research on trustworthy multimodal conversational AI for health monitoring, integrating audio, language, and clinical data with uncertainty quantification to generate reliable, context-aware responses to health-related queries.
Specialization in Artificial Intelligence and Machine Learning.
Thesis research conducted in-person at the Department of Computer Science and Technology, University of Cambridge, under the supervision of Prof. Cecilia Mascolo and Prof. Domenico Talia.
The dissertation focused on advancing machine learning models for the automation of clinical consultations and auscultation, exploring multimodal approaches, audio-based insights, and explainable AI techniques.
Thesis: API Development for IoT Applications using the Mulesoft Suite, Supervisor Prof. G. Fortino (Highly Cited Researcher in Computer Science by Clarivate for five consecutive years)
© Gaia A. Bertolino — Last update: October 2026
