Software engineer · Cape Town
Twenty years of Java and Spring Boot in financial services. Evenings and weekends go into AI tooling. Nothing on this page is a plan; it’s work that exists.

Each service comes from something I’ve built and run for myself first.
Pull-request review in your CI, running on local models so your code isn’t sent to a third-party API. Built on ReviewGate.
Hands-on sessions on getting predictable output from LLMs with slash references and repeatable prompt systems. Based on the Playbook.
Notion-to-WordPress publishing with review gates and automation, so drafts reach published without copy-paste. It’s the pipeline I publish through.
MCP servers that let AI assistants read your site or system, starting read-only. This site runs one.
Only listed once a project has been through at least three prototype iterations, each built and tested.
Splits a document into sentences or clauses and checks each one with an AI model. Plugin-based, single HTML file, live on the playground.
Code review as a CI step, using local models through Ollama instead of a hosted API. Now open source as ai-code-reviewers.
A knowledge-graph CLI with snapshot and write-ahead-log persistence and conflict detection, built to catch documentation drift.
A command-line tool that gives you the same slash commands across different AI models.
A Java and Spring Boot app for managing MCP server proxies from one place.
A TypeScript browser game with a test suite: one flight mode and two platformer worlds. Playable on the playground.
First posts are on the way: build notes from the projects above.
Slash References Playbook: Control LLMs Like A Pro
The prompting method I use every day, written up as a practical playbook.