Matteo Ruffini

Matteo Ruffini

Machine Learning Scientist

Contact Me

About Me

I am a senior machine learning scientist at Amazon, where I am working on Information Retrieval and Unbiased Learning to Rank.

Before joining Amazon, I was a PhD student at Universitat Politècnica de Catalunya, in Barcelona, where I obtained a PhD in Computer Science.


I focus my research on both, theoretical and applied machine learning.

My current area of interest is the study and the development of machine learning algorithms able to learn unbiased models from biased user-interaction data.

In the past, I worked on methods with guarantees to learn high dimensional probabilistic latent variable models, studying their applicability to standard machine learning tasks, like clustering and natural language processing.

I also worked on machine learning for healthcare, applying latent variable models to retrieve latent patient representations in large-scale healthcare data.


Generating Synthetic but Plausible Healthcare Record Datasets

L Aviñó, M. Ruffini, R. Gavaldà

KDD workshop on Machine Learning for Medicine and Healthcare, 2018.


A New Method of Moments for Latent Variable Models

M. Ruffini, M. Casanellas, R. Gavaldà

Machine Learning Journal (MLJ), 2018.

Paper - Poster

Hierarchical Methods of Moments

M. Ruffini, G. Rabusseau, B. Balle

Neural Information Processing Systems (NIPS), 2017.

Paper - Slides - Poster

Clustering Patients with Tensor Decomposition

M. Ruffini, R. Gavaldà, E. Limón

Machine Learning for Healthcare (MLHC), 2017.

Paper - Poster

Is the Brownian bridge a good noise model on the boundary of a circle?

M. Ruffini, G. Aletti

Annals of the Institute of Statistical Mathematics, 2017.


Uncalibrated view synthesis with homography interpolation

P. Fragneto, A. Fusiello, B. Rossi, L. Magri, M. Ruffini

3D Imaging, Modeling, Processing, Visualization and Transmission (3DIMPVT), 2012.



A New Method of Moments for Latent Variable Models

European Conference on Machine Learning (ECML-PKDD), 2018


Clustering Patients with Tensor Decomposition

Machine Learning for Healthcare (MLHC), 2017


Tensor Decomposition for Healthcare Analytics

Barcelona Mathematical Days (BMD), 2017



PhD - Computer Science

Universitat Politècnica de Catalunya - BarcelonaTech (2015 - 2019)

Thesis title: Learning latent variable models: efficient algorithms and applications

Supervisor: Ricard Gavaldà

Master degree - Mathematics

Università degli Studi di Milano (2009 - 2011)

Master thesis: A canonical form for Gaussian periodic processes.

Bachelor degree - Mathematics

Università degli Studi di Milano (2006 - 2009)