I build software where correctness matters.

Professionally, I develop financial infrastructure in crypto and payments, working on payment rails, banking integrations, KYC workflows, ledger systems and backend services that support the movement of money securely and reliably.

Outside of work, I enjoy building products from first principles. My current focus is Orus, a healthcare staffing platform, together with Berlin, a development platform for orchestrating modern backend applications.

I am particularly interested in distributed systems, financial engineering and software architecture.

A research harness for crypto trading signals, built so that label leakage is hard to commit and impossible to hide.

It started with a graduate machine learning exercise that trains a classifier to predict Bitcoin's direction and reports accuracy in the low nineties. That number is real arithmetic and no forecasting at all. The target was SMA10 > SMA60 measured at the same bar, and those same moving averages sat in the feature matrix, so the sign of one column reproduces the label for 100.000% of rows. A two line pandas expression scores 100% without training anything.

Closing that, plus a shuffled split on autocorrelated features, plus reporting a hyperparameter search's own best score as performance, plus a backtest that pays no fees, leaves an honest 50.1%. None of nine model families beats simply holding the asset once costs are paid.

Live dashboard · Source

What is in it:

• Purged walk-forward cross-validation with embargo, so no training fold can see across a test boundary
• Forward-looking labels that carry the time their outcome becomes known, checked mechanically rather than by eye
• Triple-barrier labelling, volatility scaled
• Cost-aware backtesting with mandatory execution lag, per-side fees and slippage
• Bayesian and grid hyperparameter search under nested validation
• Leakage detectors that run as their own CI gate, so this class of mistake cannot be merged
• A FastAPI service and a Next.js dashboard on live Coinbase feeds

Contributions are welcome, and breaking the leakage detectors is the most useful kind.

Designing and maintaining production systems involving:

• Payment rails
• Banking integrations
• Ledger creation and approvals
• Financial reconciliation
• Individual and corporate KYC
• Account provisioning

Healthcare staffing platform connecting healthcare facilities with qualified locum professionals.

Stack

• Ruby on Rails 8
• Next.js
• PostgreSQL
• Docker
• JWT authentication
• Redis
• Nginx

A local development platform designed to simplify multi-service application development using Docker.

Reliability over novelty.

Simplicity over cleverness.

Readability over complexity.

Maintainable systems outlive brilliant code.

A measurement you cannot reproduce is an opinion.

🎓 MSc Financial Engineering

WorldQuant University

🎓 BSc Mathematical Science

University of Limpopo

• Applied Data Science I (with Honors)
• Microsoft Certified Solutions Developer
• Microsoft Certified Solutions Associate
• Developing Microsoft Azure Web Services
• ASP.NET MVC Web Applications
• Programming in C#