Gaia Grosso

Teaching & Mentoring

I teach statistics and machine learning to physicists, and physics-informed modelling to machine learners. Most of my teaching happens through research supervision.

Mentoring

14students since 2023
7institutions
4 / 10undergrad / grad & MSc

Students at MIT, Harvard, Princeton, ETH Zurich, University of Padova, IFT Madrid and IFAE Barcelona. Primary research advisor unless noted; MIT undergraduates supervised through the Undergraduate Research Opportunities Program (UROP). Linked projects lead to the resulting papers.

Current

May 2026 –
Shirley X. Yuundergraduate, Princeton
Auditing fine-tuned language models via representation drift
Feb 2026 –
Francesco Montagnograduate, IFAE Barcelona
Anomaly detection in the four-lepton final state with CMS Open Data; interpretable EFT searches with contrastive latent spaces
Sep 2025 –
Cameron McNameegraduate, Harvard (co-advised)
Interpreting large models with co-activation graphs
May 2025 –
Sergio Rodriguez-Benitezgraduate, IFT Madrid
Review of the look-elsewhere effect from classical methods to machine learning (in preparation); regularization in ML-based anomaly detection
Apr 2025 –
Shelley Tonggraduate, MIT
Continuous systematic uncertainties in contrastive embeddings for LHC data; anomaly detection in the four-lepton final state with CMS Open Data

Former

Mar – Aug 2026
Ahmed Fadhlallahundergraduate, MIT
Metrics for evaluating embedding spaces for anomaly detection in the four-lepton final state with CMS Open Data
Aug – Oct 2025
Sai Sumedh R. Hindupurgraduate, Harvard
Theory of SparKer learning dynamics
Nov 2024 – Sep 2025
Giada BadaraccoMaster’s, ETH Zurich (co-advised)
Uncertainty-aware validation of generative AI
Sep 2023 – Sep 2025
Pietro Cappelligraduate, Univ. of Padova
Generative-model validation via ML goodness-of-fit tests
Oct 2023 – Apr 2025
Lana Xuundergraduate, MIT
Contrastive representations for anomaly detection
Apr 2023 – Apr 2025
Nathaniel S. Woodwardundergraduate, MIT
Product-manifold machine learning for particle physics
2023 – 2024
Nicolò Laigraduate, Univ. of Padova
Signal-agnostic NPLM search in the CMS dimuon final state
Oct 2023 – Sep 2024
Kyle S. MetzgerMaster’s, ETH Zurich (co-advised)
Representation learning for new physics searches in CMS
Jan – Jun 2024
Jessica PrendiMaster’s, ETH Zurich (co-advised)
Model-agnostic searches in the CMS 40 MHz scouting system

Lectures and tutorials at schools

PHYSTAT School of Statistics upcoming
Lecturer · Damp, Germany · 22 February – 1 March 2027
Lecture topic to be announced.
Statistical and ML tools for anomaly detection in fundamental physics
Lecturer · First PHYSTAT School of Statistics, “Statistics in the Era of Machine Learning”, Nikhef · November 2025
A survey of classical hypothesis testing alongside density estimation, representation learning, and weak supervision — with emphasis on robustness, calibration, and uncertainty quantification.
PHYSTAT
Physics-guided optimizations: inform and inspire AI models with physics principles
Lecturer · IAIFI Summer School, Harvard University · August 2025
Encoding physical structure into models through losses, regularization, and inductive biases; and how physical principles — annealing, diffusion, entropy, energy-based models — inspire modern optimization and architecture design.
Slides (PDF) School page
Diffusion models for physics — hands-on exercises
Tutorial developer and facilitator · IAIFI Summer School, MIT · August 2024
Designed Jupyter notebooks on diffusion models, variational autoencoders, and validation methods, and led the hands-on session. Freely reusable.
Notebooks on GitHub

University courses

Same or different? Classical and ML methods for two-sample testing in science
CDS DS 595, AI Methods for Science · Boston University · March 2026
Two-sample hypothesis testing for scientific discovery: classical statistics, kernel MMD, and classifier-based tests, with applications to high-energy physics and AI model validation.
Slides (PDF)
Statistical tests for anomaly detection and scientific discovery
8.16/8.316, Data Science in Physics · MIT · April 2025
From classical hypothesis-testing frameworks to methods based on density estimation, representation learning, and weak supervision.
Laboratory of Computational Physics (Mod. A)
Teaching assistant · M.Sc. in Physics of Data, University of Padova · 2022–2023
Hands-on computational methods for a master’s programme at the interface of physics and data science.