Main Conference Day One - GMT (Greenwich Mean Time, GMTZ)
What are the strategic developments in large language models in the current industry? How is GenAI specifically being used in top institutions? Perspectives from investment banking and fund management.
- Theodora Lau - Founder, Unconventional Ventures
- Gary Kazantsev - Head of Quant Technology Strategy, Bloomberg
- Nicole Königstein - Chief AI Officer, Head of AI & Quant Research, quantmate
- Rama Cont - Professor of Mathematics and Chair of Mathematical Finance, University of Oxford
- Dan Nechita - Head of Cabinet for MEP, European Parliament
- Stefano Pasquali - Head of Investment AI Modelling & Research team, BlackRock
- Manually entering complex trades for what-if analysis is a slow and unreliable component of what is otherwise a highly effective process of trade execution
- We discuss the key techniques involved in building LLM-based trade entry from natural language trade description contained in unstructured email messages or threads
- We present specialized prompting and RAG techniques for dealing with complex trade definition formats and logic
- GPT-4o, LLAMA 2/3 and Mixtral models are compared on examples from all major asset classes
- Open-source code for all examples and tests will be provided
- Alexander Sokol - Executive Chairman, CompatibL
How can institutions plan for long term technological investments and decisions? What is the necessary infrastructure required for this? Where is quantum computing and other similar technologies today?
- Joachim Mnich - Director for Research and Computing, CERN
AMD Instinct processors are designed to revolutionize data center performance by accelerating deep learning, artificial neural networks, and high-performance computing (HPC) applications. Built on the AMD CDNA architecture, these processors offer advanced Matrix Core Technologies and support a wide range of precision capabilities, from INT8 and FP8 for AI to FP64 for HPC. The latest MI300 series combines the power of AMD Instinct accelerators with AMD EPYC processors, providing enhanced efficiency, flexibility, and programmability. This presentation will explore the key features and benefits of AMD Instinct processors, highlighting their role in advancing AI and HPC workloads.
- Michael Fernandez - Head of GPU Business Development for Financial Applications, AMD
A deep look from one of the architects of the EU’s AI act into its implications, capital markets, geopolitics, and a year of decisive elections across half the planet.
- Helyette Geman - Research Professor & Member, John Hopkins University & Board of Bloomberg Commodity Index
Managing investments and trading for quant finance alongside regulatory requirements and reputational risk.
- Richard Turner - Managing Director, Currency Management, Mesirow Financial
- Ying Poikonen - Executive Director, Head of Modelling Group EMEA Region, SMBC Group
- Andrea Macrina - Professor of Mathematics, University College London
- Edward Baker - Advisor, Tipping Frontier
During this workshop we will focus on practical techniques for building reliable and effective LLM-based workflows for extracting accurate information from natural language documents or unstructured data.
Unlike chat and creative writing applications where a great deal of variation in responses to the same query is acceptable but sometimes even required, implementing business processes using LLMs demands stable and reproducible results that can be rigorously tested. This presents a unique set of engineering challenges due to the inherent randomness of LLMs.
During the workshop will discuss how to overcome these challenges to build effective LLM-based workflows with hands-on examples from trading and risk.
- Alexander Sokol - Executive Chairman, CompatibL
Make sure to visit this year's PhD posters in the Foyer and vote for your top 3 posters on the ConnectMe app!
- Hamza Bodor - PhD Candidate, Université Paris 1 Panthéon-Sorbonne
- Natascha Hey - PhD Student, Ecole Polytechnique, Paris
- Hoang Dung Nguyen - PhD Student, Université Paris Cité
- Adil Rengim Cetingoz - PhD Student, University of Paris I: Panthéon-Sorbonne, Paris
- Sturmius Tuschmann - PhD Candidate, Imperial College London
- Robert Boyce - PhD Student, Imperial College London
- Edward Selig - MSc Graduate, Imperial College London
- Valeria Varlashova - MSc Graduate, Imperial College London
- Ruben Kerkhofs - PhD Student, University of Leuven
- Nathan De Carvalho - PhD Student, Université Paris Cité
- Shaun (Xiaoyuan) Li - PhD Student, Université Paris 1 Panthéon-Sorbonne
- Herve Andres - PhD Student, Ecole nationale des ponts et chaussées
- Gabin Taibi - PhD Student, University of Twente
- Santiago Walliser - PhD Student, University of Zurich
It is well known that yield curves have low effective dimensionality, and can be accurately represented using very few latent variables. The recent extension to nonlinear representations by means of autoencoders (AE) provided further improvement in accuracy compared to the classic linear representations from the Nelson-Siegel family, or those obtained using principal component analysis (PCA). In this presentation, we use AE to capture the historical dependence structure of interest rates, and introduce risk-neutral forward rate dynamics that are consistent with a given AE curve manifold.
We first derive a general condition for the AE-based forward-rate curve to admit a no-arbitrage evolution. Then, by allowing a small convexity-driven deviation from the AE curve manifold, we derive a risk-neutral modeling framework that is arbitrage-free and incorporates the information built into the AE (low-dimensional) curve manifold. The numerical results we showcase are based on historical market swap data for multiple currencies.
- Andrei Lyashenko - Head of Market Risk and Pricing Models, Quantitative Risk Management (QRM), Inc.
- Fabio Mercurio - Global Head of Quant Analytics, Bloomberg L.P.
- Alexander Sokol - Executive Chairman, CompatibL
Distributionally robust optimization studies the worst deviation of an evaluation functional in the neighborhood of some given model of interest. We derive explicit sensitivity analysis under no arbitrage and calibration constraints. In particular, this provides an original first order hedge against model risk.
- Nizar Touzi - Chair of the Department of Finance and Risk Engineering, New York University
The demand for equity portfolios that minimize risk and volatility while still delivering attractive returns is more pressing than ever. Designing quantitative strategies that operate at very low beta (<0.1) is essential for buffering downside risk and achieving stable, all-terrain performance. By leveraging active trading dynamic beta machines and focused single-stock management, Quantmade addresses this challenge.
- Michael Geke - CEO, Quantmade
- Matthew Rooney - VP, Head of Quant Analytics Europe, Selby Jennings
Empower alpha generation with Yukka’s news AI, transforming unstructured data into actionable insights.
- LLM-Driven Market Insights: Discover how our AI-powered chat facilitates real-time news exploration and extraction of critical insights.
- Alpha-Focused Data: Learn how our unique data set identifies news events and provides quantitative insights critical to financial markets.
- Signal Creation Simplified: Dive into a practical example of building powerful quantitative investment signals, step by step.
- Flexible Applications: Explore a variety of formats — from interactive talking avatars to tailored reports — designed to make large volumes of unstructured news data easily digestible.
Join us to see news AI in action for quant investing.
As factor investing becomes more popular in Credit markets we dive into the Fidelity Multifactor Credit Model. We look at how factors work in credit markets, ways to assess their efficacy in a complex markets and the novel way we approach credit curve construction to improve relative value metrics.
- Joe Hanmer - Global Head of Quant, Fidelity International
- Mihail Turlakov - Quant Trader, Independent
We design and train machine learning models to capture the nonlinear interactions between financial market dynamics and high-frequency trading (HFT) activity. In doing so, we introduce new metrics to identify liquidity-demanding and -supplying HFT strategies. Both types of HFT strategies increase activity in response to information events and decrease it when trading speed is restricted, with liquidity-supplying strategies demonstrating greater responsiveness. Liquidity-demanding HFT is positively linked with latency arbitrage opportunities, whereas liquidity-supplying HFT is negatively related, aligning with theoretical expectations. Our metrics have implications for understanding the information acquisition process in financial markets.
- Gbenga Ibikunle - Professor and Chair of Finance, University of Edinburgh
How to integrate Artificial Intelligence with the Human Experience in a complete investment process journey.
Experience and past performance and future developments in a rapidly changing investment industry
- Daniele Bernardi - CEO, Diaman Partners
The ACE short rate model—originally developed by Gregory Pelts and now rephrased, refined, and made more accessible by Matthias Heymann—is the first to combine all of the most desirable analytical properties in one interest rate framework: It is low-dimensional (with any dimension n≠2), complete (modeling all tenors), arbitrage-free, highly flexible (offering 2n+1 discrete parameters plus the functional noise parameter σ(x,t)), time homogeneous if desired, and it imposes a lower bound on rates. Furthermore, it has the rare property of being unspanned (i.e., its bond price function does not depend on σ), which greatly facilitates calibration to the full set of swaptions in practice. Recently, the speaker has extended the model to cover the rates of multiple currencies, as well as their FX spot and forward rates.
- Matthias Heymann - Former Market Data Scientist, Millennium
- Andrew McClelland - Director, Quantitative Research, Numerix
- Leon Tatevossian - Adjunct Professor, New York University
- Reminder on the path-dependent volatility (PDV) model of Guyon and Lekeufack (2023), where the instantaneous volatility is a linear combination of recent trend and recent realized volatility.
- Adding a parameter that unlocks enough volatility on the upside to reproduce S&P 500 and VIX smiles.
- This PDV model, motivated by empirical studies, comes with computational challenges, especially in relation to VIX options pricing and calibration. We propose an accurate neural network approximation of the VIX which leverages on the Markovianity of the 4-factor version of the model. VIX is learned as a function of the Markovian factors and the model parameters.
- We use this approximation to tackle the joint calibration of S&P 500 and VIX options.
- Julien Guyon - Professor of Applied Mathematics, École nationale des ponts et chaussées
We propose a holistic theory of XVA based on recognition of all real-life derivatives as liabilities or portfolios thereof
- Andrey Chirikhin - Credit Quantitative Analytics, Barclays Investment Bank
- Recent episodes and basic concepts
- From entity to system stress testing
- Banking system applications
- Financial system applications
- Jérôme Henry - Principal Adviser – DG Macroprudential Policy and Financial Stability, European Central Bank
Recent Term Premium values given by yield-curve based models are unintuitive and do not seem to reflect the current environment. Survey data are surely the gold standard for Term Premium estimates but they are infrequent. We show that most yield curve based models may be represented with simple linear combinations of key rates, and by combining this property with survey data, we arrive at simple, daily frequency term premium valuations for the US and EU zones.
- Jessica James - Managing Director, Senior Quantitative Researcher, Commerzbank AG
We discuss the challenges of regime changes in risk management, including some notes on rough volatility and interest rate curves.
- Marcos Costa Santos Carreira - Head of Equity Quants, XP Inc
In this presentation, we will more broadly explore the relationship between inflation, growth and macro assets. We'll also be examining more broadly the relationship between inflation and growth across multiple countries. We'll look at how forecasts for inflation and growth can be used to create systematic trading rules for macro assets such as commodities, EMFX, etc.
- Saeed Amen - Founder, Turnleaf Analytics
One of the main challenges facing the factor investing field is the proliferation of unreliable empirical methods suffering from specification biases and hardly applicable to all asset classes. In this seminal paper, we introduce a unified theoretical framework for factor investing that stems from the structural relationship between equities and bonds. We build on the Langstaff and Schwartz (1995) structural model to design a tractable method for generation of optimal factor portfolios of corporate bonds. We then suggest a stepwise calibration procedure that helps implement our model in a tractable and a computationally efficient way.
- Hamza Bahaji - Head of Financial Engineering and Investment Solutions, Amundi ETF, Indexing & Smart Beta,, Amundi
We introduce rhetorical engineering for client portfolio construction using MLM. Rhetorical engineering generalises and structures prompt engineering. We adapt classical rhetoric based on identity (ethos), emotion (pathos), and intention (logos) to communicate client identity and intentions. We demonstrate that this identity, emotion, and intention materially (i.e. statistically significantly, allowing for multiple tests) modifies how the client, as reflected by MLM, views the strengths, weaknesses, opportunities, and threats of companies. As an example we apply rhetorical engineering to SWOT scoring of all companies in the S&P500(c), and show how the company selection for portfolio inclusion materially changes. We consider identity by investor gender and investment horizon. For emotion we consider happy, angry and calm-and similar states induced in MLM. For intention we consider investors excited by AI/ML, worried about future pandemics, and concerned about sustainability. The company selection also changes the out of sample portfolio performance. Thus rhetorical engineering is a highly adapted and flexible tool for communicating client / investor identity and preferences for portfolio construction.
- Chris Kenyon - Global Head of Quant Innovation, MUFG Securities
During this session, we explore the asset turnover of a strategy that combines multiple trading signals within the FactSet's Programmatic Environment (FPE). We discuss constructing signal portfolios, focusing on their temporal cross-sectional autocorrelation and its impact on turnover. By standardizing portfolios and assuming specific weights distributions, we derive a formula to estimate turnover: the turnover depends on the first-order autocorrelation of the portfolio representation of the signal. Subsequently, we present an optimization process for combining multiple signals that maximizes the information ratio of the resulting composite signal while incorporating a turnover control mechanism. This framework is validated through backtest simulations on the Russell 1000 universe. The results empirically demonstrate the effectiveness of our approach to reduce turnover by managing autocorrelation.
- Georgi Mitov - Director, Quantitative Research, FactSet
- Matthew Rooney - VP, Head of Quant Analytics Europe, Selby Jennings
- Jim Gatheral - Presidential Professor of Mathematics, Baruch College, CUNY
- Aitor Muguruza Gonzalez - Chief Artificial Intelligence Officer, Kaiju Capital Management
- James McGreevy - Quantitative Researcher, Kaiju Capital Management
- Žan Žurič - Quantitative Researcher, Kaiju
I will review my thinking and some research about
- Adaptive and inelastic markets
- Short-maturity vol arbitrage - implied from realised volatility, smile expansion, and SSVI
- Long-maturity vol arbitrage - hedged Monte Carlo and generative paths calibration
- Mihail Turlakov - Quant Trader, Independent
- An achievable lower bound for a Bermudan is not always the max European
- Finding a robust (model-free) lower bound for Bermudans is surprisingly difficult
- It is important for designing “Bermudan discount” adjustment models prevalent in the industry for the last 30+ years
- We prove that generalized call spread is THE (achievable) lower bound
- Vladimir Piterbarg - Managing Director, Head of Quantitative Analytics & Development, NatWest Markets
- Mihail Turlakov - Quant Trader, Independent
- Toby Hill - Associate Director - Quantitative Research & Trading, Selby Jennings
- Hamza Bodor - PhD Candidate, Université Paris 1 Panthéon-Sorbonne
- Hamza Bahaji - Head of Financial Engineering and Investment Solutions, Amundi ETF, Indexing & Smart Beta,, Amundi