hi, I am Matteo.

I am a research scientist in Machine Learning and Neuromorphic Computing. I am fascinated by life and intelligence, and I like to study complex systems. I love to play music and dance. Most importantly, I am a human being going through this life with the help of human connections and open communication.

I usually experiment and do stuff with LLMs. I also work with "efficient and exotic hardware" that implement neural computations with analog circuits. I have a PhD in Neuroinformatics with a thesis on synaptic plasticity and predictive processes in biological and artificial networks.

projects.

finetuning an LLM on trippy mathematical reasoning (wih a couple of RTX 2080 cards)

2025 - A (playful?) experiment finetuning a language model to channel “high on LSD”, dreamy rationales, push finetuning to the limit, and see if the model finds useful rationale patterns in trippy chain of thoughts.
"The underlying structures of self-attention: symmetry, directionality, and emergent dynamics in Transformer training"

2025 - We show that bidirectional and autoregressive training induces symmetric and directional self-attention matrices, and that these structures can be leveraged for improving performances.
programming mixed-signal devices on-chip with feedback control

2024 - I develop a novel feedback control algorithm for on-chip training of mixed-signal neuromorphic chips, opening doors for applications in smart sensors, IoT, and autonomous systems.

last update: