Leonardo Cella | Mathematics, AI & Energy Trading

About me

I work at the intersection of mathematics, machine learning, and energy trading. Since September 2026, I have been Chief Scientist Officer (CSO) at an energy trading startup now taking shape.

I also joined AIPIA’s Technical and Scientific Committee in September 2026.

September 2026

A new chapter in energy trading

I am excited to help build this venture alongside highly experienced colleagues with outstanding track records in the industry.

We are building a trading approach grounded entirely in AI and quantitative methods, including AI agents. Our primary focus is the Nordic power markets, with ambitions that extend beyond imbalance trading. For me, this is a demanding and exhilarating challenge. Let the new adventure begin!

A heartfelt thank you to Duferco Energia and the colleagues who made my time there so rewarding.

Research interests

I use mathematics to model and understand the physical world. I am interested in how real-world events unfold, how their patterns can be captured in models, and how those models can inform decisions under uncertainty.

In energy markets, this means connecting physical dynamics to price behaviour and trading decisions through statistics, machine learning, and reinforcement learning.

Background

Before this venture, I worked at Duferco Energia. Earlier, I spent nearly two years in the Energy Portfolio team at Axpo Italia, focusing on customer consumption and the intraday power market.

At the Italian Institute of Technology, I worked with Massimiliano Pontil on multi-task learning, meta-learning, and stochastic linear bandits. This research also involved Karim Lounici. I collaborated with Claudio Gentile, then at Google in New York, on non-stationary stochastic bandits.

I hold a PhD from the University of Milan, where I was advised by Nicolò Cesa-Bianchi. From July to October 2018, I was an Applied Scientist intern in Amazon’s Music ML team in Berlin, mentored by Giovanni Zappella.

Since 2018, I have served on programme committees for machine learning conferences including AISTATS, NeurIPS, and ICML.

Selected publications

Multi-task Representation Learning with Stochastic Linear Bandits L. Cella, K. Lounici, G. Pacreau, and M. Pontil. International Conference on Artificial Intelligence and Statistics (AISTATS), 2023.
Group Meritocratic Fairness in Linear Contextual Bandits R. Grazzi, A. Akhavan, J. I. T. Falk, L. Cella, and M. Pontil. Conference on Neural Information Processing Systems (NeurIPS), 2022.
Multi-Task and Meta-Learning with Sparse Linear Bandits L. Cella and M. Pontil. Conference on Uncertainty in Artificial Intelligence (UAI), 2021.
Best Model Identification: a Rested Bandit Formulation L. Cella, C. Gentile, and M. Pontil. International Conference on Machine Learning (ICML), 2021.
Meta-learning with Stochastic Linear Bandits L. Cella, A. Lazaric, and M. Pontil. International Conference on Machine Learning (ICML), 2020.
Stochastic Bandits with Delay-Dependent Payoffs L. Cella and N. Cesa-Bianchi. International Conference on Artificial Intelligence and Statistics (AISTATS), 2020.
Efficient Linear Bandits through Matrix Sketching I. Kuzborskij, L. Cella, and N. Cesa-Bianchi. International Conference on Artificial Intelligence and Statistics (AISTATS), 2019.
Deriving item features relevance from past user interactions L. Cella, S. Cereda, M. Quadrana, and P. Cremonesi. Conference on User Modeling, Adaptation and Personalization (UMAP), 2017.

News

September 2026. I joined an energy trading startup at its inception as Chief Scientist Officer (CSO).
September 2026. I joined the AIPIA Technical and Scientific Committee.
July 2024. I joined Duferco Energia as a Quant Trader.
October 2022. I moved to industry, joining the Energy Portfolio team at Axpo Italia.
May 2022. I attended Re-thinking High-dimensional Mathematical Statistics at the Mathematisches Forschungsinstitut Oberwolfach.
December 2021. I attended the Meeting in Mathematical Statistics at CIRM in Marseille.
December 2021. I was recognised as an outstanding reviewer at NeurIPS 2021.
July 2021. I presented Best Model Identification: a Rested Bandit Formulation at ICML, joint work with Massimiliano Pontil and Claudio Gentile.
July 2021. I presented Multi-Task and Meta-Learning with Sparse Linear Bandits at UAI, joint work with Massimiliano Pontil.
October 2020. I joined the Computational Statistics and Machine Learning lab at IIT in Genoa, working with Massimiliano Pontil on meta-learning and stochastic bandits.
June 2020. I presented Meta-learning with Stochastic Linear Bandits at ICML, joint work with Alessandro Lazaric and Massimiliano Pontil.
June 2020. I presented Stochastic Bandits with Delay-Dependent Payoffs at AISTATS, a project inspired by my experience at Amazon.
September 2019. I gave a talk on Stochastic Bandits with Delay-Dependent Payoffs at the MAPLE Workshop in Milan.
April 2019. I presented Efficient Linear Bandits through Matrix Sketching at AISTATS in Okinawa, joint work with Ilja Kuzborskij and Nicolò Cesa-Bianchi.
July–October 2018. I was an Applied Scientist intern at Amazon in Berlin.

Other interests

I care deeply about combating human trafficking and reducing barriers to education, particularly for children. I support these causes through my involvement with Cooperativa Cielo and the Sant’Egidio community in Genoa.