Founding Members (Est. 2005)
Mark Raffaelli, CFA® FRM®
Mark started his career in financial markets in London as a spot FX, Options, and Swap Inter-Dealer Broker (IDB). Many in South Africa know him from his foundational work at the Bond Exchange of South Africa (BESA) before he co-founded Geometric Progression. At BESA, his key contributions included the authoring the ILB & FRN pricing specifications and key yield curve methodologies still anchoring the market, chairing the Quant Committee, spearheading Capital Adequacy reforms, new trading rules and BESA’s VaR model.
This deep market structure experience, combined with a successful history in ML/DL, now drives his design of Geometric's advanced quantitative solutions, including optimisation engines for portfolio allocation and security selection. Mark has frequently appeared on television and in newspapers, is often a guest speaker at industry conferences, and has served as a guest lecturer for post-graduate studies on bond markets. Committed to education, he has also taught advanced mathematics to high schoolers aiming for top grades.
Sandra Raffaelli
Sandra's unique background combines the dynamism of guiding backpacker tours in the iconic Kruger National Park and its surrounds with years of corporate precision in client operations at Deloitte. This diverse experience forged her strong governance and client focus, essential to Geometric Progression.
The Rest of the Crew
Beyond our founding members, Geometric Progression thrives on a dynamic team of bright minds. We regularly welcome talented Honours and Masters graduates and provide opportunities for promising undergraduates in Mathematics, AI, and Computer Science. These individuals are integral to our innovative capacity, contributing fresh perspectives to our cutting-edge quantitative solutions and agentic AI development.
What We’ve Built
-
Agentic AI & Intelligent Automation Hubs
Leveraging agentic AI for transformative automation, we've successfully built containerized code development hubs with distinct agents (code manager, backend/frontend developers, unit testers, evaluator) for streamlined software creation. Additionally, autonomous agents perform periodic sentiment analysis on portfolio stocks, while collaborative agent teams conduct in-depth stock research, portfolio allocation, selection, and optimization. -
Data Foundation for Success
Our projects begin with robust data cleaning and feature engineering: meticulous cleaning, stationarity adjustments preserving signals, rigorous statistical checks (correlation, independence), and advanced transformations, ensuring model reliability. -
Time Series Analysis and Prediction
We have a strong record in forecasting market dynamics. We apply classical models (e.g., ARIMA, ARIMAX, SARIMAX, VARMA) as baselines and compare the results against advanced machine learning/deep learning models (LSTMs, N-BEATS, TimeGPT etc.). We provide full predictive distributions and key point estimates for robust decision-making. -
Yield Curves and Interest Rate Modelling & Prediction
Our expertise encompasses the full spectrum of yield curve analytics, from constructing consistent curves using sophisticated interpolation to advanced modelling and prediction of their future dynamics. We forecast shifts in level, slope, and curvature, providing critical insights for fixed-income strategy, risk management, and derivative pricing. -
Security Analysis and Index Intelligence
Successful projects include dynamic security classification and factor-based selection. For fixed income, we have engineered systems for automated bond index tracking and rebalancing. Furthermore, we have developed predictive models for bond index reconstitutions, forecasting changes in constituents and weightings to enable proactive portfolio positioning. -
Advanced Portfolio Optimization
We employ traditional (Mean-Variance) and later optimization methods (Black-Litterman, Risk Parity, CVaR). A key focus is bond portfolio optimization, where strategies include precise duration matching while systematically tilting portfolios for enhanced convexity. All optimization incorporates fair allocation principles, rigorously adhering to client mandates, liquidity constraints, and regulatory requirements. -
Market Resilience and Risk Management
We build regime change detection models (e.g., Hidden Markov Models) to help strategies adapt to shifting market conditions. Our comprehensive risk management solutions include Value-at-Risk (VaR) modeling, stress testing, and the development of specialized dynamic interest rate hedging programs. -
Specialized Solutions for Financial Institutions
For banking clients, we have successfully developed systems for market abuse detection, identifying patterns such as spoofing and front-running, including those executed by high-frequency trading algorithms. We also design custom factor models and performance attribution systems.
