Generative Methods in Quant Finance
Quant teams need reliable ways to generate scenarios, fill data gaps, and create synthetic market data for stress testing, model validation, and trading research. The approach must be transparent, data‑efficient, and fast enough for real use—without turning every task into a deep‑learning science project.
Gaussian Mixture Models (GMMs, and their lognormal counterparts, LGMMs) offer that balance: they fit high‑dimensional distributions well, provide closed‑form conditionals for quick sampling, and can incorporate practical enhancements. The result is a simple simulation recipe that uses only Gaussians and uniforms while remaining auditable.
Join our special guest A/Prof. Dr. Joerg Kienitz of m|rig on September 4, 2025 at 10am EDT as he shows how GMMs/LGMMs power backcasting of missing time series, generation of yield curve scenarios across currencies, and conditional implied volatility surfaces – while reducing static arbitrage and preserving covariance structure.
A/Prof. Dr. Kienitz will cover the following topics, utilizing Python notebooks to illustrate examples:
- Why mixtures: when GMMs/LGMMs beat heavy generative models
- Fitting & sampling in practice
- Conditional simulation for scenarios, imputation, and backcasting
- Synthetic yield curves and implied vol surfaces with no‑arbitrage checks/repair
Featured Speakers
Joerg Kienitz, PhD, Adjunct Associate Professor
Joerg Kienitz has been a finance professional for over 25 years and has held numerous positions as Head of Quantitative Analytics at firms including Postbank/Deutsche Bank, Deloitte and LSEG Post Trade Solutions. Currently, he is the Director of Quantitative Methods at m|rig, a Frankfurt-based consultancy. Joerg is primarily involved in consulting on model validation, model development, and model implementation across all asset classes, but with a focus on counterparty credit risk and interest rates. He also advises on the application of Machine Learning for Quantitative Finance. Moreover, Joerg is a trainer for public and onsite classes.
As an academic, Joerg supervises MSc and PhD students at the universities of Wuppertal (BUW) as a lecturer and Cape Town (UCT) as an Adjunct Associate Professor. His research topics include advanced financial modelling, numerical methods, stochastics and machine/statistical learning. Further to his academic engagements, he speaks at well-known quantitative finance conferences including QuantMinds International and WBS Quantitative Finance Conference.
Joerg holds a PhD in stochastic analysis and probability theory and has authored several papers in well-known journals such as Quantitative Finance, Journal of Computational Finance and Risk. He has also written four books: “Monte Carlo Frameworks: Building Customisable High Performance C++ Applications” (with Daniel J. Duffy) and “Financial Modelling: Theory, Implementation and Practice with MATLAB Source” (with Daniel Wetterau), published by Wiley Finance; and “Interest Rate Derivatives Explained, Volume I” and “Interest Rate Derivatives Explained, Volume II” (with Peter Caspers), published by Palgrave Macmillan.
Greg Murray
Greg Murray is responsible for increasing awareness of the Numerix brand in financial markets around the globe and contributing to Numerix’s strategic growth initiatives. Previously, he oversaw product and field marketing initiatives at the company, and he started his tenure in a sales role. Prior to Numerix, Mr. Murray worked in derivative analytics sales roles at other software firms, and he held derivative trading positions for seven years as an option market-maker and proprietary trader across a variety of asset classes.
