• @ Berkeley
• bachelor in Econometrics @ Bocconi
• part of Hephaestus Applied Artificial Intelligence
• part of Bocconi AI lab (department of Computing Science)

• 🏆 'YC Hackaton': Winners of the 2026 YC Hackaton on MCP applications.
• 🏆 'Cyberwave Hackaton': Winner for the scrapegraphai track (https://devpost.com/software/robothelper).
• 🏆 'B@B Hacks Hackaton': Top 5 as best project overall
• 🚀 'Codex Hackaton': 2026 Codex Hackaton project
• 🧙‍♂️ 'OpenClaw wrapper': a simpler and dumber version of openclaw with better UX

🧙‍♂️ Lately I made a new SOTA result on the PENS @ 'Refactor'

🌦️ 'Weather derivatives': I was into finance

🧩 With Braico and Filo we explored ARC–AGI with Reinforcement Learning:

• 🌤 Weather Derivatives: Climate modeling with Ornstein-Uhlenbeck processes using Wavelet Networks.
• ARC (v1): Custom DSL and action space with Wolpertinger Actor-Critic for large combinatorial domains.
• ARC (v2): Added learnable action embeddings and ViT state encoding with a pretrained world model.
• ARC (v3): World modeling.
• 'Combinatorially complex action spaces'

🧠 Under the CS department @ Bocconi we are working on:

• 'RL World Models': Benchmarking LeJEPA+VICReg vs. ViT encoder–decoder world models in RL environments.