CV

PhD Student in Generative AI

Delft University of Technology

I am PhD student in Computer Science at TUDelft advised by Prof. Justin Dauwels and Prof. Geert Leus. I am currently working on Generative Models for video prediction and model based RL downstream tasks. I have practical experience in: 1) Object-Centric Representation Learning, Unsupervised Representation Learning, Deep Information Theory, Disentanglement Learning. 2) Generative Modelling: VAEs, GANs, Diffusion Models, Autoregressive Transformers, Spatio-Temporal generative models (I.e, ConvLSTM, VQVAE+TransformerXL, etc.), Modular Neural Networks (I.e., Neural Production Systems, Recurrent Independent Mechanisms). 3) Model-based Reinforcement Learning: Multi Agent RL, PPO-based models, Recurrent State Space Models (RSSMs, such as Dreamer), Transformer-based world models (I.e., STORM, IRIS, Transdreamer), Hierarchical models (I.e., Director).

• Video Prediction Modelling
• Model-based Reinforcement Learning
• Vision-Language Modelling
• Representation Learning
• Object-Centric Representation Learning

Research Intern

Mila - Quebec AI Institue

Supervised by Prof. Yoshua Bengio and Anirudh Goyal (DeepMind). I worked on three projects:

• Model-based Reinforcement Learning
• Hierarchical World Models
• Object-Centric World Models

MSc

Delft University of Technology

BioRobotics

BSc

Polytechnic University of Turin

Mechanical Engineering

SURF Compute Grant

SURF

Compute resources to conduct top-notch ML research.

Infineon’s IPCEI PhD Booster Scholarship

Infineon Technologies

A 3 years scholarship to travel, attend conferences and world wide events.

MILA Internship Scholarship

MILA Quebec AI Institute, Montreal, Quebec

Research Internship supervised by Prof. Yoshua Bengio and Dr. Anirudh Goyal

TUDelft Honor Program

Delft University of Technology

Top 2% students

Young Talent Award

Politecnico di Torino

Top 2% students

Extreme Precipitation Nowcasting using Transformer-based Generative Models

ICLR 2024 Workshop: Tackling Climate Change with Machine Learning

This paper presents an innovative approach to extreme precipitation nowcasting by employing Transformer-based generative models, namely NowcastingGPT with Extreme Value Loss (EVL) regularization

AlphaTC-VAE: On the relationship between Disentanglement and Diversity

The Twelfth International Conference on Learning Representations (ICLR 2024)

In this work, we introduce AlphaTC-VAE, a variational autoencoder optimized using a novel total correlation (TC) lower bound that maximizes disentanglement and latent variables informativeness.

Nowcasting of Extreme Precipitation Using Deep Generative Models

ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

In this paper, novel deep generative models are proposed for precipitation nowcasting. These models are equipped with extreme-value losses to more reliably predict extreme precipitation events.

Stateful active facilitator: Coordination and environmental heterogeneity in cooperative multi-agent reinforcement learning

The Eleventh International Conference on Learning Representations (ICLR 2023)

In this work we formalize the notions of coordination level and heterogeneity level of an environment and present HECOGrid, a suite of multi-agent RL environments that facilitates empirical evaluation of different MARL approaches across different levels of coordination and environmental heterogeneity by providing a quantitative control over coordination and heterogeneity levels of the environment. Further, we propose a Centralized Training Decentralized Execution learning approach called Stateful Active Facilitator (SAF) that enables agents to work efficiently in high-coordination and high-heterogeneity environments through a differentiable and shared knowledge source used during training and dynamic selection from a shared pool of policies.

Adaptation through prediction: Multisensory active inference torque control

IEEE Transactions on Cognitive and Developmental Systems

Adaptation to external and internal changes is of major importance for robotic systems in uncertain environments. Here, we present a novel multisensory active inference (AIF) torque controller for industrial arms that shows how prediction can be used to resolve adaptation.

Active inference in robotics and artificial agents: Survey and challenges

ArXiv

Survey and challengese of Active Inference approachs in robotics and artificial intelligence.

Multimodal-VAE Active Inference Controller

2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

We present a novel active inference torque controller for industrial arms that maintains the adaptive characteristics of previous proprioceptive approaches but also enables large-scale multimodal integration (e.g., raw images).