Professional Experience
Machine Learning Researcher
- Started a collaboration of professors, researchers, and companies to experiment new roadmaps to reach AGI.
- Trained a GPT2-like architecture from scratch and experimented with variations.
- Built a self-improving LLM agent that leverages the scientific method and experiments to build a model of the world.
- Invented a recurrent architecture for LLMs with power law information storage behaviour and tunable exponent.
- Used Alpha-Zero like Monte Carlo tree search to improve the decoding performance of Language Models.
Management Consultant
- Freed 50+M€ from the reserves of 70 Italian banks by inventing and implementing a data pipeline.
- Automatized 100+ hours of work by writing Python scripts.
- Responsible for the work of 3 experienced consultants and the weekly reports to client's management.
- Contributed to the definition of the strategy for 10+ companies in transportations, energy and banking.
Education
Scuola Normale Superiore (SNS)
Best Italian university for Physics. Full scholarship is granted for the about 10 physicists selected every year.