Last update 25/09/2026

Studying

• Mechanistic Interpretability, and in general explainability for LLM and ML models.
• Statistics
• AI Safety - Started BlueDot technical AI safety course
• CyberSecurity

Building

• Exploration of the differences and similarities between DiffusionGemma and its AutoRegressive twin.
• While not being even an average chess player, having fun with models that play chess (not necessarily LLMs) and how to change their behaviour from outside.
• Funny enough, a widget for Apple Watch that is completely useless, inspired by the magic item that was the Seiko TV Watch

Working

• AI Engineer @ pillar.