Selected work

This is not a complete portfolio. It is a sample of the kind of work I take on, and of the kind of problem I used to own inside a bank.

Economic scenario generation for insurers

I built a stochastic economic scenario model for long-term insurance valuation and risk analysis. Its C++ engine jointly simulates interest rates, inflation and asset returns across multiple markets, calibrating them to market data while preserving the relationships between economic variables. It produces internally consistent scenarios that insurers can use to explore how assets and liabilities may behave across a wide range of future economic conditions.

XVA calculation for derivatives portfolios

I built an XVA analytics engine that allows structurers to assemble derivatives portfolios in Excel or Python, run Monte Carlo simulations and calculate valuation adjustments including CVA, DVA, FVA and KVA. By making these calculations available during portfolio construction, the tool allows users to assess the effects of counterparty credit risk, funding and capital before a proposed structure is traded.

Digital credit approvals

I led a three-year programme that used BCBS 239/RDARR as the foundation for digitising high-volume, vanilla credit approvals in a Corporate and Investment Bank. The Digital Credit Paper replaced a largely paper-based process—hard copies passed between reviewers and marked up in red pen—with an automated workflow that cut turnaround times by almost two-thirds. It was built entirely on a new cloud technology stack while complying with stringent data-governance and privacy requirements.

Quantitative risk, in production

At Standard Bank I was responsible for teams that owned Basel rating models (PD, LGD, EAD), trading-book exposure modelling (Monte Carlo, some XVA), and risk data (BCBS 239 / RDARR) for the Corporate and Investment Bank. The part of that job I still like is getting strong quantitative people off local MATLAB, Excel and SQL and onto a pipeline that builds the model in code and deploys it — in that environment, C# and Python onto Azure.

Applied generative AI

I focus on the aspects of generative AI that have already demonstrated value in large organisations, rather than chasing the bleeding edge or the surrounding hype. That means choosing bounded use cases, applying proven techniques, and building the data, controls and operating processes needed for reliable use in a corporate environment. The objective is a useful capability that can be adopted and governed—not an impressive demonstration that cannot survive production.