📢 News
Read the blog post about my thesis: Efficient SAM2
Small updates about my academic, professional, and personal life.
2026
🔐 HEAT makes nonlinearity approximation iterations learnable, so weights and approximations co-adapt — 3.1× fewer iterations, 1.6× fewer bootstraps and 1.4× faster encrypted GPT-2 inference.
🎸 Have you ever wondered who is the most… among your friends! Play and settle it!
🇧🇷 From the 23rd to the 27th of Aprile, I’ll be in Rio de Janeiro to present my paper MASS!
🎤 I’ll be participating as an invited speaker at Cybersecurity — Una Gara da Vincere 2026, talking about Homomorphic Encryption in Machine Learning Models!
😌 Ours Zero-Shot Quantization via Weight-Space Arithmetic shows that PTQ robustness is a transferable direction in weight space, enabling zero-shot quantization robustness transfer across ViT models.
📰 Our paper MASS, MoErging through Adaptive Subspace Selection, has been accepted at ICLR 2026.
2025
👨🏻💻 Today is my first day as a PhD student at the GLADIA research group, advised by Prof. Emanuele Rodolà!
🥳 I defended my Master’s thesis titled “Efficient SAM2” at Sapienza University of Rome. With a final grade of 110/110 cum laude, I am now officially a computer scientist!
🎤 On the 16th of September 2025, I will be presenting my paper MASS at “Breaking the Monolith” workshop at ICIAP 2025!
📍 End of my Erasmus. I’m moving back to Rome 🇮🇹 to pursue my PhD at Sapienza.
🎓 I defended my Master’s thesis at TUM, Efficient SAM2 (read the blog post🤩).
🥳 I’ve been accepted as a PhD candidate in informatics at Sapienza University of Rome! Resulting in the fourth highest score, I was awarded a funded position. I will be advised by professor Emanuele Rodolà.
📰 My first paper MASS, MoErging through Adaptive Subspace Selection is available on arXiv!
2024
📍 My Erasmus is about to start! I will move to Munich 🇩🇪, where I will be studying and doing my thesis research at TUM for the entire 2024/2025 academic year.