Abstract: Optimal transport has emerged as a powerful mathematical framework with applications in machine learning, economics, and finance. However, scaling optimal transport to large datasets across distributed systems remains a significant challenge, especially when the data is spread out across multiple agents which cannot communicate the raw data. This talk will present Federated Sinkhorn, a distributed algorithm for computing optimal transport efficiently in decentralised settings. By leveraging distributed computing principles, the Federated Sinkhorn algorithm enables scalable and privacy-preserving solutions to high-dimensional transport problems. We will discuss the methodology, highlight key results, and explore potential applications in financial modelling.
13:00 - 14:00
Will Shoosmith, Quantum and AI Innovation Manager, HSBC
Lecture 2: Innovation at HSBC: Round Peg in a Hexagon Hole
Abstract: This talk will explore how HSBC is using advanced computing technologies, including quantum computing and artificial intelligence, to shape the future of financial services. It will outline the bank's approach to innovation, including how opportunities are sourced, validated, and scaled through internal R&D and external collaboration. Additionally, the talk will address the cultural, organisational, and technical shifts that financial institutions require to keep up with emerging technologies.
14:00 - 15:30 - Lunch break
15:30 - 16:30
Dr. Giulio Giaconi, Senior AI Research Scientist, HSBC
Lecture 3: Four Moment Stochastic Processes for Flexible Uncertainty Modelling in Bayesian Regression
Abstract: Bayesian optimization has established itself as one of the most effective black-box optimization frameworks for problems where evaluations are expensive, noisy, or derivative-free, ranging from hyper-parameter tuning in deep learning to automated materials discovery. This talk offers first an introduction to the topic that bridges foundational concepts with recent applications in finance. Additionally, we present a new stochastic process for regression that generalizes the Gaussian process to four moments. This provides for a more flexible capacity to model uncertainty and tail behavior that is often present in multiple domains for which precise characterization of tail behavior is particularly critical, such as in the case of financial markets and when assessing the structural integrity of degrading engineering equipment. We extend the literature describing the flexible four moment distribution called shifted generalized lognormal distribution to model stochastic processes, and derive the form of the process as well as their standard computations in applications.
Speaker bios:
- Jeremy Kulcsar is an Advanced Compute Research Scientist at HSBC. He works on both theoretical topics related to classical compute and AI, and their way of practical implementation for banking use cases. Within the bank, he already published about distributed computing for optimal transport, and his current research interests lie in machine learning and causality. However, he is also occasionally working on quantum-inspired topics. Before joining HSBC, Jeremy worked in the area of Data Science for 5 years, notably for investment banks and insurance companies. He holds a BSc in Applied Physics from Universite Paris-Saclay, an MSc in Computer Science from Ecole Centrale Paris and a Master’s in Management from ESSEC Business School.
- Will Shoosmith is a Quantum Innovation Manager at HSBC. He is at the forefront of investigating quantum and advanced computing technologies for financial institutions. He contributes to various research projects, such as with quantum entropy sources, scalable optimisation methods, and innovative neural network architectures, while also supporting outreach and strategic partnership development. Previously, as a HSBC Technology Graduate, he gained expertise in AI and cloud engineering, building LLM-based tools and testing quantum-enhanced algorithms for financial optimisation. Will holds a BSc in Biochemistry from Imperial College London, specialising in computational-omics, with research focused on cryo-EM image analysis using advanced alignment and reconstruction algorithms, as well as network-based approaches to microbial enzyme function analysis. His passion lies in the intersection of biology, computation, and next-generation technologies.
- Dr. Giulio Giaconi is an Advanced Compute Scientist at HSBC. Dr. Giulio Giaconi received his B.Sc. and M.Sc. degrees (honours) in Communications Engineering from Sapienza University of Rome, Italy; and his Ph.D. degree in information-theoretic privacy in smart meters from the Department of Electrical and Electronic Engineering, Imperial College London. He is currently a Senior Applied Research Scientist at HSBC in the Emerging Technology, Innovation, and Ventures team. His current research interests lie in the areas of optimization and machine learning for finance applications. Previously, he was a Senior Data Scientist at Ofcom, in the Data Innovation Hub; and a Senior Research Scientist at BT Applied Research, in the Security Futures Practice. Giulio was awarded best PhD thesis on information theory in the UK and Ireland by the IEEE UK and Ireland Information Theory Chapter, and the Excellent Graduate Student Award from Sapienza University of Rome.
