Paolo Conti

Paolo Conti

PhD student in Scientific Machine Learning

Politecnico di Milano

I am a PhD student in scientific machine learning at Polytechnic University of Milan. I am fascinated by the immense world that arises from blending core numerical methods with artificial intelligence learning techniques, which is what I am focusing on in my research activities. Lifelong learning is my attitude, and working day-by-day for long-term goals is my methodology. Mathematical engineer by formation, aerobic gymnast by passion.

Interests
  • Scientific Machine Learning
  • Reduced Order Modeling
  • Multi-fidelity data fusion
Education
  • PhD in Scientific Machine Learning (in progress)

    Politecnico di Milano, University of Washington

  • MSc in Mathematical Engineering, 2021

    Politecnico di Milano, Sorbonne University

  • BSc in Mathematical Engineering, 2018

    Politecnico di Milano

Experience

 
 
 
 
 
Imperial College London
Visting Researcher
December 2023 – January 2024 London, UK

Led a research project to develop generative AI/ML frameworks for data-driven, reduced-order modeling under uncertainty.

  • Conceptualized and implemented a data-driven framework based on variational reduced-order modeling with variational dynamics identification for scientific discovery in the presence of model and measurement uncertainties.
 
 
 
 
 
SimTech - Cluster of Excellence
Visting Researcher
October 2023 – November 2023 University of Stuttgart, Germany
Developed the Python package VINDy to perform data-driven modeling of dynamical systems with generative AI.
 
 
 
 
 
Artificial Intelligence Institute in Dynamic Systems
Research Intern
October 2022 – June 2023 University of Washington, Seattle (US)

Developed ML algorithms for dimensionality reduction and system identification in complex, high-dimensional scenarios in engineering sciences.

  • Designed and constructed physics-informed models for Micro-Electrical Mechanical Systems (MEMS) devices. Application and validation on MEMS micromirrors and resonator devices.
  • Developed a multi-fidelity method to recover and predict high-quality solutions from multi-modal, low-fidelity data sources.
 
 
 
 
 
Applied Mathematics, Sorbonne University
Exchange program
September 2019 – July 2020 Paris, France
Study abroad coursework in the departments of Applied Mathematics and Sorbonne Polytech.
 
 
 
 
 
International Gymnastics Federation
Professional Athlete
International Gymnastics Federation
January 2010 – December 2021 Bergamo, Italy
  • Member of the National Team of Aerobic Gymnastics from 2010 to 2021.
  • World medallist and European champion.
  • Experiences in coaching and coreographing in Italy, France, Finland, Hungary, Lithuania and United States.

Publications

(2024). VENI, VINDy, VICI: a variational reduced-order modeling framework with uncertainty quantification. arXiv.

PDF Cite Code DOI

(2024). EKF-SINDy: Empowering the extended Kalman filter with sparse identification of nonlinear dynamics. CMAME.

PDF Cite Code Project DOI

(2024). Multi-fidelity reduced-order surrogate modeling. Proceedings of Royal Society A.

PDF Cite Code DOI

(2023). Reduced order modeling of parametrized systems through autoencoders and SINDy approach: continuation of periodic solutions. CMAME.

PDF Cite DOI

(2023). Multi-fidelity surrogate modeling using long short-term memory networks. CMAME.

PDF Cite DOI

Projects

*
Physical generative AI
An interpretable data-driven framework for building generative reduced order models with embedded uncertainty quantification
Physical generative AI
EKF-SINDy digital twin
Coupling data assimilation with system identification to build a data-driven digital twin
EKF-SINDy digital twin
Modeling from measurements
Data-driven methods for model discovery.
Modeling from measurements
What do you need to climb the charts?
Analysis of a Spotify dataset.
What do you need to climb the charts?

Contact