I am a senior research scientist at Google DeepMind working on reinforcement learning and alignment of language models to human feedback. I am also highly involved in the post-training of Gemma models.
Prior, I was a PostDoc at ETH Zürich, working with Andreas Krause and Nicolò Cesa-Bianchi.
My research has focused on sequential decision-making in different settings and application areas, with a particular focus on active learning and sample-efficiency.
I am also very interested in the interplay between game theory and machine learning.
In September 2022, I obtained a Ph.D. from ETH Zürich advised by Maryam Kamgarpour and Andreas Krause, where I studied learning and efficiency in multi-agent systems [pdf].
I have interned with Google Brain in 2022 and Google Research in 2021 working with large language models,
the Institute for Transport Planning and Systems of ETH Zürich in 2018, and the United Technologies Research Center (Cork, Ireland) in 2016.
I obtained a MSc. in Robotics, Systems and Control from ETH Zürich (2017) and a BSc. in Automation Engineering from Politecnico di Milano (2015), both with honors.
At ETH, I have been Teaching Assistant of the Master's level courses Control Systems II (Head TA) and System Identification during 2018-2021.
(for an updated list, see Google scholar)
DockGame: Cooperative Games for Multimeric Rigid Protein Docking
V. R. Somnath*, P. G. Sessa*, M. R. Martinez, A. Krause
Preprint, 2023.
[pdf]
Adversarial Causal Bayesian Optimization
S. Sussex, P. G. Sessa, A. Makarova, A. Krause
International Conference on Learning Representations (ICLR), 2024.
[pdf]
Distributionally Robust Model-based Reinforcement Learning with Large
State Spaces
S.S. Ramesh, P. G. Sessa, Y. Hu, A. Krause, I. Bogunovic
International Conference on Artificial Intelligence and Statistics (AISTATS), 2024.
[pdf]
Multitask Learning with No Regret: from Improved Confidence Bounds to Active Learning
P. G. Sessa*, P. Laforgue*, N. Cesa-Bianchi, A. Krause
Neural Information Processing Systems (NeurIPS), 2023 (to appear).
[pdf]
How Bad is Selfish Driving?
Bounding the Inefficiency of Equilibria in Urban Driving Games
A. Zanardi*, P. G. Sessa*, N. Kaeslin, S. Bolognani, A. Censi, E. Frazzoli
IEEE Robotics and Automation Letters (RA-L), 2023.
[pdf]
Movement Penalized Bayesian Optimization with Application to Wind Energy Systems
S.S. Ramesh, P. G. Sessa, A. Krause, I. Bogunovic
Neural Information Processing Systems (NeurIPS), 2022.
[pdf]
Efficient Model-based Multi-agent Reinforcement Learning via Optimistic Equilibrium Computation
P. G. Sessa, M. Kamgarpour*, A. Krause*
International Conference on Machine Learning (ICML), 2022.
[pdf]
[bib]
Boosting Search Engines with Interactive Agents
L. Adolphs, B. Boerschinger, C. Buck, M. C. Huebscher, M. Ciaramita, L. Espeholt, T. Hofmann, Y. Kilcher, S. Rothe, P. G. Sessa, L. Sestorain
Transactions on Machine Learning Research, 2022.
[pdf]
Online Submodular Resource Allocation with Applications to Rebalancing Shared Mobility Systems
P. G. Sessa, I. Bogunovic, A. Krause, M. Kamgarpour
International Conference on Machine Learning (ICML), 2021.
[pdf]
[bib]
[poster]
Contextual Games: Multi-Agent Learning with Side Information
P. G. Sessa, I. Bogunovic, A. Krause, M. Kamgarpour
Neural Information Processing Systems (NeurIPS), 2020.
[pdf]
[bib]
[poster]
Learning to Play Sequential Games versus Unknown Opponents
P. G. Sessa, I. Bogunovic, M. Kamgarpour, A. Krause
Neural Information Processing Systems (NeurIPS), 2020.
[pdf]
[bib]
[poster]
(presented at ICML 2020 Workshop: Real World Experiment Design and Active Learning)
No-Regret Learning from Partially Observed Data in Repeated Auctions
O. Karaca*, P. G. Sessa*, A. Leidi, M. Kamgarpour
International Federation of Automatic Control (IFAC) World Congress, 2020.
[pdf]
[bib]
Mixed Strategies for Robust Optimization of Unknown Objectives
P. G. Sessa, I. Bogunovic, M. Kamgarpour, A. Krause
International Conference on Artificial Intelligence and Statistics (AISTATS), 2020.
[pdf]
[bib]
[video]
No-Regret Learning in Unknown Games with Correlated Payoffs
P. G. Sessa, I. Bogunovic, M. Kamgarpour, A. Krause
Neural Information Processing Systems (NeurIPS), 2019.
[pdf]
[bib]
[poster]
[video]
Bounding Inefficiency of Equilibria in Continuous Action Games using Submodularity and Curvature
P. G. Sessa, M. Kamgarpour, A. Krause
International Conference on Artificial Intelligence and Statistics (AISTATS), 2019.
[pdf]
[bib]
[poster]
(Contributed talk at NeurIPS 2018 Workshop: Smooth Games, Optimization and Machine Learning)
Designing Coalition-Proof Reverse Auctions over Continuous Goods
O. Karaca, P. G. Sessa, N. Walton, M. Kamgarpour
IEEE Transactions on Automatic Control (TAC), 2019.
[pdf]
[bib]
A hybrid stochastic approach for offline train trajectory reconstruction
P. G. Sessa, V. De Martinis, A. Bomhauer–Beins, U. Weidmann, F. Corman
Public Transport, 2020.
[pdf]
[bib]
Filtering approaches for online train motion estimation with onboard power measurements
P. G. Sessa, V. De Martinis, F. Corman
Computer-Aided Civil and Infrastructure Engineering, 2019.
[pdf]
[bib]
Exploiting structure of chance constrained programs via submodularity
D. Frick*, P. G. Sessa*, T. A. Wood, M. Kamgarpour
Automatica, 2019.
[pdf]
[bib]
A Hybrid Dynamic-Kinematic EKF for Train Trajectory Estimation
P. G. Sessa, V. De Martinis, F. Corman
International Conference on Intelligent Transportation Systems (ITSC), 2018.
[pdf]
[bib]
From Uncertainty Data to Robust Policies for Temporal Logic Planning
P. G. Sessa, D. Frick, T. A. Wood, M. Kamgarpour
International Conference on Hybrid Systems: Computation and Control (HSCC), 2018.
[pdf]
[bib]
Exploring the Vickrey-Clarke-Groves Mechanism for Electricity Markets
P. G. Sessa, N. Walton, M. Kamgarpour
International Federation of Automatic Control (IFAC) World Congress, 2017.
[pdf]
[bib]
* denotes equal contribution