I am a Visiting Researcher at MIT, advised by Pierfrancesco Beneventano and Tomaso Poggio, working on agentic systems, fundamental limits of AI, and theorems about training neural networks.

I am writing this after years of building and questioning models: before MIT, I worked on foundation models at Logitech's CTO office with Jonathan Dan; at Idiap Research Institute, Philip N. Garner supervised my research on bio inspired text to speech and spiking neural networks.

I am still shaped by a winding education: engineering physics at Politecnico di Milano, then nuclear engineering through a joint master's across ETH Zurich, EPFL, and PSI. I did not complete that degree; when my interests pulled elsewhere, I moved into data science and computational neuroscience at EPFL, where the questions finally lined up with the work I wanted to pursue.

I am always glad to hear from others. If you want to collaborate or trade ideas, coffee optional, say hi; I learn best from other people's notebooks.