[Meet our faculty] Giulio Franzese, assistant professor in the data science department at EURECOM

Q1. What is your academic trajectory?

GF. I hold a Master’s Degree in Telecommunications Engineering (2016) and a Ph.D. in Electronics and Telecommunications Engineering (2021) from Politecnico di Torino. My academic path began with a focus on statistical signal processing, which expanded during my internship at the German Aerospace Center in Munich. My Ph.D. research at Politecnico di Torino focused on efficient machine learning methods, laying the groundwork for my shift toward more advanced methodological research in machine learning. During my Ph.D., I was a visiting scientist at EURECOM in 2019, which marked my first encounter with the Data Science Department. After completing my Ph.D., I returned to EURECOM as a post-doctoral researcher in 2020. Over these years, I developed strong expertise in generative models, multimodal information theory, and machine perception. This led to multiple research projects, where I explored innovative areas in multimodal learning and machine perception.

In 2024, I was appointed as an Assistant Professor in the Data Science Department at EURECOM, where I continue to build on my expertise in machine learning, particularly in the fields of diffusion-based generative models and multimodal information theory.

Q2. What is the expertise you bring to the Data Science department?

GF. My expertise lies in advancing the theoretical foundations of machine learning, with a focus on multimodal learning, generative modeling, and information theory. One of the core areas I am passionate about is understanding how systems process and integrate multiple sources of information across different modalities — such as text, image, and audio — and how we can improve the performance of machine perception systems. I am also deeply involved in exploring representation learning and its potential for extracting meaningful insights from complex, high-dimensional data.

In terms of practical applications, I have been engaged in developing methods for machine perception that are grounded in rigorous mathematical frameworks, while also making them applicable to real-world systems, such as those in autonomous systems and communications. Through my research, I aim to contribute to the department by enhancing its focus on methodological work. I hope to push the boundaries of how we can refine machine learning algorithms, making them more robust and adaptable to the complex multimodal scenarios that define the modern digital landscape​.

Q3. What made you choose to come to EURECOM?

GF. I was first drawn to EURECOM due to its reputation for pioneering work in the fields of Data Science and Machine Learning. During my time as a visiting scientist in 2019, I was impressed not only by the depth of research but also by the collaborative and interdisciplinary environment. EURECOM fosters a culture where theoretical research thrives alongside practical, real-world applications, and this is exactly the environment I was looking for to develop my own work.

Returning as a post-doctoral researcher and now as an Assistant Professor, I have found EURECOM to be a place that encourages innovation and exploration. The Data Science Department offers a unique space where cutting-edge research is actively supported, with ample opportunities for collaboration across multiple domains. This has allowed me to pursue my passion for advancing machine learning methodologies while contributing to impactful projects that span academic and industrial research. It is this dynamic environment that solidified my decision to continue my career here​.

Q4. What are your future goals and if you had to sketch a five-year plan for your research goals, what would that be?

Over the next five years, my research will focus on advancing key areas within multimodal learning, information theory, and generative modeling. I aim to explore how we can better model complex interactions between different data types and modalities, improving both the efficiency and scalability of machine learning systems. A major goal is to develop robust methods for representation learning, particularly in contexts where integrating multiple sources of information is essential — such as autonomous systems, communications, and healthcare applications.

In parallel, I want to push the boundaries of mutual information theory to improve our understanding of how to quantify and leverage information from multiple modalities. This theoretical work will have broad implications, potentially transforming how we design models that can adapt to the diversity of real-world data.

Additionally, I aim to focus on cross-domain applications where machine learning can provide meaningful insights, such as in the development of systems that perceive and interact with their environment in an intelligent, adaptive manner. Building a research team around these core areas is another key goal, where I envision close collaboration between PhD students and postdoctoral researchers to develop innovative methodologies that balance theory and practice. In essence, my five-year plan revolves around creating a robust theoretical framework for machine learning that can be seamlessly applied to real-world, multimodal problems.