Learning from Human Feedback
Learning from human preferences, demonstrations, and other feedback.
Research
Research interests
Efficient Sample Use and Reuse
Convergence of actor-only and actor–critic methods with efficient data reuse.
Stability in Reinforcement Learning
Bridging control theory and RL through stability and mixing properties.
Structured Sequential Decision-Making
Exploiting structure in multi-armed bandits and Markov decision processes.
Research projects
iBeCHANGE
Addressing psychosocial and lifestyle risk factors for primary cancer prevention
European Union · Horizon Europe RIA · Grant 101136840
FAIR
Future Artificial Intelligence Research, Spoke 4: Adaptive AI
MUR / NextGenerationEU · PNRR Extended Partnership · PE00000013
CLINT
Climate Intelligence: extreme events detection, attribution, and adaptation
European Union · Horizon 2020 · Grant 101003876
Software and datasets
Gym4ReaL ↗
Open-source Gymnasium suite for benchmarking reinforcement-learning algorithms on realistic environments.
Python · Apache-2.0AReS ↗
Open-source patient simulator for testing automated control strategies for total intravenous anesthesia.
Python & MATLAB · MITARLO ↗
Python library and extensible framework for automated reinforcement learning pipelines.
Python · MITSelected recognitions
2026
ELLIS ScholarOutstanding junior scientists recognition
2024
Two spotlight papers at ICMLTop 3.5% of submissions
2023
Oral presentation at ICMLTop 2.39% of submissions
2023
Notable paper at AISTATSTop 1.9% of submissions
2021
Spotlight presentation at NeurIPSTop 2.9% of submissions
2021
Premio NeoDottori di Ricerca Marco CadoliAIxIA · Best Italian Ph.D. thesis in AI
2018
Oral presentation at NeurIPSTop 0.62% of submissions
Technology transfer
I am co-founder of ML cube S.r.l., a Politecnico di Milano spin-off, part of Kayrhos Group, providing cutting-edge solutions for machine learning systems and life-cycle-management optimization.