Biography
Umberto Michelucci studied Theoretical Physics in Italy, the United States, Germany and England, focusing on simulations and the theory of high temperature superconductors. He obtained his PhD in Machine Learning applied to Physics from the Portsmouth University in England. He worked in various industries for many years, gaining industrial experience in various fields. He also founded and directed the AI Center of Excellence at Helsana Versicherung AG in Zürich. He is also the co-founder of the company TOELT LLC, that focuses on large international research project with a strong AI component. He is working often with Google and NVIDIA on various project. He has many years of teaching at various levels, Bachelor, Master, PhD and for continuing education. He also holds a Postgraduate Certificate in Higher Education. He is Professor at HSLU. He is also the co-founder of the HSLU Applied AI Center.
He has written six books on Deep Learning, machine learning, statistics and AI in industry with Springer Nature. My Springerlink books: https://link.springer.com/search?contributor=Umberto+Michelucci&sortBy=newestFirst.
Personal Website: https://umbertomichelucci.me/
Associations / Membershipts
Education
PhD in Machine Learning applied to Physics
Master in Theoretical Physics
Postgraduate Certificate in Higher Education
Awards
Interviews
Research
My main interests lie in Applied & Computational Mathematics with focus on Scientific Machine Learning, Inverse Problems & Uncertainty Quantification.
Inverse problems & scientific estimation (recover latent scientific quantities from indirect/noisy measurements; study identifiability, stability, and consistency, with spectroscopy, sensing, medical/remote imaging as testbeds).
Uncertainty quantification for reliable ML (metrics and validation approaches that explicitly incorporate label/measurement noise; bootstrap and related tools for realistic performance assessment).
Fundamental limits & model-agnostic bounds.
Algorithms to estimate Bayes error / intrinsic task difficulty.
Learning in function spaces (learning & Statistics in Infinite-Dimensional Spaces).
Operator/function-valued learning and generalization on spectra, curves, and fields.
Structure- and domain-informed ML (explainability with physics/chemistry priors).
Domain adaptation and interpretable models that map spectral features to physico-chemical mechanisms; small-datasets.
Scientific sensing & instrumentation with ML (ML-enhanced, often multitask, sensors; super-resolution for fast spectroscopy; end-to-end design and analysis).
Open benchmarks/datasets for satellite vision and medical imaging to make evaluation reproducible and failure modes visible.
Effects of label/input/algorithmic noise and model misspecification; design of estimators and training schemes that are stable to perturbations and measurement error.
Functional Theory and Random Matrix Theory and their relations to the mathematical foundations of machine learning.
Computational astrophysics. I have an on-going cooperation with INAF (Istituto Nazionale Astrofisica, Italy) in Bologna and Milano. Projects are on scientifically informed algorithm development, uncertainty and error estimation of deep learning algorithms applied to astrophysics, spectral model-agnostic feature extraction, etc.
Computational food technology. I am working on algorithms to extract chemical information from various food types (as maize, olive oil, wine, etc.) from optical measurements, in particular fluorescence and Raman spectroscopy (in collaboration with Prof. Dr. Venturini). I am also interested in developing explainable ML approaches, to study if algorithms really learns from the chemistry of the samples, or from other aspects that are not related to physics or chemistry.
Talks and Conferences
2nd Workshop on Machine Learning in Infinite Dimensions - Sept. 2025 ETH Zürich - Poster "Infinite-Dimensional Nature of Spectroscopy and Why Models Succeed, Fail and Mislead"
Talk @ Spital Emmental (AIMED) - August 2025 - "Mehrwert und Umgang mit KI im Alltag"
Talk @ Italian Astrophysics National Institute (INAF) (Bologna) - Jan 2025 - "Machine Learning in Science: Opportunities, Challenges, and New Frontiers Making Science, not pseudo-science"
Talk @ Politecnico di Torino (POLITO) (Turin, Italy) - Jan 2025 - "Artificial Intelligence Projects in Bioengineering and Medicine Generating Impact"
Invited Lecture @ Uni St. Gallen, (St. Gallen) - Jan 2025 - "Road to become an AI Company Challenges and chances for SMEs"
Talk @ FFHS (Zürich) - Jan 2025 - "Generative KI und Digitalisierung in Bildung und Forschung: Herausforderungen, Chancen und ethische Überlegungen"
Talk @ Virtuial AI Summit (Online, Berlin) - "The AI-Ready Company: Strategies for Future-Proofing Your Business"
Talk @ European Space Agency (ESA) (Madrid) - April 2024 - "Precision Recognition: Training Deep Learning Algorithms for Multiscale Vehicle Detection in Satellite Imagery"
Summer Schools
Masters and PhDs
Selected Masters Tutored
2026-Present
(Topics are in the definition phase)
1) Kevin Kurinjirappalli - Topic: Responsible Use of AI in Education
2) Fabio Hänni - Topic: Governance und Risikomanagement von AI-Agenten in der Vermögensverwaltung
3) Ingold Pascal Remo - Topic: study of Peptides and proteins with machine learning / collaboration with Politecnico di Torino (Italy)
4) Sarankan Maheswaran - Topic: RAG system for historical documents
5) Thivvirthan Krishnakumar - Topic: data engineering for medicine (to be defined) / collaboration with Children Hospital Zürich
2025-2023
1) Joshua Hügli, "3D Reconstruction of Trapezium Bone from 2D X-Rays", 2025
2) Dr. Nathalie Alexander, "Fractal Dimension Analysis of Pediatric Brain Structures: Developmental Curves Across Age" 2025.
The thesis has resulted in a paper: Alexander, N., Gucciardi, A., & Michelucci, U. (2026). Quantifying segmentation-driven measurement error in infant brain MRI and its impact on volume and fractal dimension reliability. European Journal of Radiology Artificial Intelligence, 100087.
3) Marta Martinozzi, "Automatic Anomaly Detection in sensor data for ICU patients" (in collaboration with the Emmental Hospital), 2025.
4) Sarp Koc, "Drone Aerial Imaging Analysis in Smart Agriculture", 2025
5) Morteza Kiani Haftlang, "Computer Vision Framework for 3D Cancer Detection: Comprehensive assessment of self-supervised deep learning architectures", 2024
The thesis has resulted in a paper: Haftlang, M. K., Malmir, M., Parand, F., Michelucci, U., & Ghazouali, S. E. (2025). Barlow-Swin: Toward a novel siamese-based segmentation architecture using Swin-Transformers. arXiv preprint arXiv:2509.06885.
6) Matthias De Paolis, "Deep Learning-based Detection of Photovoltaic Installations: A Case Study in the Canton of Zurich", 2024
6) Julia Netzel (now Software Developer at Google), "Statistical Stability of Super-Resolution for Astronomical Imaging", 2023 (ETH)
7) Christian Schmid, "Value Generation with Computer Vision Techniques in the Olive Oil Industry - a Knowledge Graph Framework", 2023
Selected PhD Students Tutored
1) Arnaud Gucciardi, "Brain Symmetry and Structure in Children MRI Imaging", 2025
2) Dr. Michela Sperti, "Predicting Personalized Pathological Risks and Dynamics in Cardiology How to Exploit Data-Driven Approaches in Clinical Decision Support Systems", (collaboration with the Politecnico di Torino, Italy) 2025
