Research

Learning from systems whose important events are rare, complex or stochastic.

My research background combines physics-based simulation, numerical algorithms and data analysis—from atomistic rare events to DNA interactions and neuromorphic materials.

Go to publications

PhD research · 2021–2022

Atomistic Simulations, IIT

Research period in Michele Parrinello’s group, focused on atomistic simulations, enhanced sampling and data-driven descriptions of slow molecular processes.

Finding the slow modes behind rare events

Molecular simulations often cannot reach the timescales of infrequent but important events. Enhanced-sampling methods address this limitation by accelerating a small set of collective variables that capture the slow processes governing a transition.

Transfer-operator methods and machine learning provide a route to identify these slow modes from simulation trajectories. This methodology formed the scientific context of my PhD research period and connected statistical mechanics with modern data-driven modelling.

Methodological context

Deep-TICA for rare-event sampling

The Parrinello-group paper combines neural-network transfer operator eigenfunctions with enhanced sampling to learn effective collective variables from biased trajectories.

This paper is included as context for the group’s methodology; I am not one of its authors.

Bonati, Piccini & Parrinello, PNAS (2021)

01 Generate trajectories Collect configurations from an initial simulation.
02 Learn slow modes Analyse time-lagged dynamics with transfer operators.
03 Build collective variables Represent the processes that limit convergence.
04 Accelerate sampling Promote transitions and reconstruct the landscape.

Peer-reviewed work

Selected publications

Two collaborations in which I contributed directly to software, algorithms, simulations and scientific analysis.

Diagram of a three-dimensional resistor network with nodes, links and conductances — Network model from Mambretti et al., Scientific Reports 12, 12234 (2022), licensed under CC BY 4.0.

Scientific Reports · 2022

Dynamical stochastic simulation of complex electrical behaviour in neuromorphic networks of metallic nanojunctions

We developed a large three-dimensional resistor-network model with non-linear conduction and stochastic conductance updates. At sufficient scale, the model qualitatively reproduces electrical behaviours observed in nanostructured gold films and supports their study as neuromorphic systems.

My contribution

Co-conception of the algorithm; development, implementation and testing; simulations, data processing and analysis.

Read the publication

Concept diagram showing DNA strands as competing species and resources in an artificial ecosystem — DNA ecosystem concept from Mambretti et al., Entropy 24, 458 (2022), licensed under CC BY 4.0.

Entropy · 2022

OxDNA to Study Species Interactions

We used coarse-grained molecular dynamics to model single-stranded DNA oligomers as species competing for resource strands. Simulations and experiments explored how sequence overlap and three-dimensional structure influence competitive advantage.

My contribution

Software development, methodology and investigation, including the molecular simulations used to study competitive and cooperative interactions.

Read the publication

Education

A foundation in physics

2019–2021

M.Sc. in Physics

University of Milan · 110/110 cum laude

2015–2018

B.Sc. in Physics

University of Milan · 110/110 cum laude

2021–2022

PhD research period

Italian Institute of Technology · Former PhD candidate

Continue exploring

Research is one part of the story.

See how this foundation now informs applied R&D, computer vision and system engineering.

View experience Download CV Email me