Marco Lorenzo and Georgiana Petroiu

A2A

Abstract: In today's dynamic energy markets, timely and precise decision-making remains a critical challenge. This presentation explores a system that employs advanced artificial intelligence (AI) techniques to convert raw news data into structured insights for energy trading. The approach integrates real-time data acquisition, natural language processing (NLP), and machine learning models, including fine-tuned GPT algorithms, to filter, evaluate, and classify news items based on their relevance to market conditions. The system architecture—comprising stages for data preparation, containerized processing, and detailed analytics—is designed to mirror the judgment of human analysts while managing a high volume of information. By automating the classification and scoring of news, the framework supports more informed market decisions and may enhance trading strategies. The seminar will briefly outline the system’s design and discuss its potential and limitations, encouraging a reflective dialogue on the evolving role of AI in risk management and quantitative finance.

Fabrizio Lillo

University of Bologna and Scuola Normale Superiore - Pisa

Abstract: Optimal execution of large orders and optimal trading are classical problems in quantitative finance and of great importance for the financial industry. Most existing models make simplifying assumptions in order to achieve analytical tractability, most notably that the trading environment is constant. I will present some recent advancements obtained with Reinforcement Learning to optimal execution in a time varying liquidity environment and to optimal trading when the signal predictability is not constant. Then, I will consider the case of multiple RL agents simultaneously performing an optimal execution and I will show that the strategies learned by the agents deviate significantly from the Nash equilibrium of the corresponding market impact game.

Flavia Rufini

Intesa Sanpaolo

Abstract: In recent years we have seen an increasing use of Generative AI models in our society. At Intesa Sanpaolo, we are particularly concerned with the implementation of Responsible AI practices and compliance with the European regulation on AI systems (AI Act). Developing guardrail frameworks is crucial to mitigate risks arising from the use of Generative AI: generation of false information (hallucinations), production of harmful or discriminatory content, vulnerability to prompt injection, disclosure of sensitive information. The key point on the development of this framework is to find the optimal trade-off between standardization and customization to ensure flexible adaptiveness to bank rules and policies. In this way we can take advantage of the potential of LLM in a safe environment. 

Christian Kappen

d-fine

Abstract: In this talk, I will explain how Monte Carlo methods can be used to simulate complex interdependent systems over long time horizons. I will discuss simulations of optimal asset control, market risk hedges, counterparty risk mitigants, and capital buffers, both from a mathematical and a technical perspective, with applications in the finance and the energy sector.

Vulkan Nir

Oxford University

Sara Pomponi and Edoardo Schiavo

Cassa Depositi e Prestiti

Abstract: The world is experiencing a period characterized by enormous social, political, and technological changes, which are reflected in the financial markets through unprecedented levels of uncertainty and volatility. This environment calls for the timely implementation of effective hedging strategies to manage evolving risks. We analyze some examples of hedges using financial derivatives from various asset classes, focusing on practical aspects related to the design, structuring and implementation of these hedges, and discuss the challenges arising from their treatment under current accounting principles.

Andrea Pallavicini

Intesa Sanpaolo

Abstract: Financial products with early-termination features allow investors to participate in the performance of underlying assets while granting them the flexibility to exit their positions before the maturity date. This presentation explores regression-based Monte Carlo methods to price and calculate sensitivities for these products. We will progress from standard polynomial regressions to advanced algorithms, including randomized feed-forward neural networks, randomized recurrent neural networks, and signature-based methods.