Experimental Projects 🚧
Here are some of the ongoing and exploratory projects I’m working on:
ArchiWeb
A collaborative, browser-based ArchiMate® 3.2 modelling tool, modelled on Archi. It keeps Archi’s metamodel, notation, rules and file formats, and adds real-time multi-user editing.
- Technologies: TypeScript, Yjs (CRDTs), SQLite, Git, Model Context Protocol (MCP)
- Features: Live collaboration with presence and per-user undo, Git sync compatible with Archi’s coArchi2, views catalog, model health dashboard,
.archimateand Open Group Exchange import/export - AI: An MCP server lets assistants such as Claude explore and edit models live, including impact analysis and health checks
Cadence
A cycling route planner that runs entirely in the browser. Click the map to plan a ride, or let it generate a loop.
- Technologies: JavaScript (no build step), OpenStreetMap, BRouter, Open-Meteo
- Features: Loop generator with difficulty levels, climb detection, elevation profile, surface and road-type breakdown, wind and weather matched to departure time, GPX import/export
- Links: Try it live · Source on GitHub
Browser-based Policy Q&A Agent
A lightweight Q&A system for internal policy documents, running locally with open-source LLMs and ChromaDB.
- Technologies: ChromaDB, Python, Mini LLM, Ubuntu Server Pro
- Features: Fast querying, private deployment, CI/CD integration
Architecture-as-Code CI/CD Toolkit
Toolkit to turn architecture practices into programmable, repeatable pipelines.
- Tools: Wardley Maps, Markdown, GitHub Actions
- Goal: Improve architecture traceability and continuous validation
Local LLM Playground
Experimenting with different open-source LLMs and comparing performance on local machines.
- Focus: Fine-tuning, latency testing, local inferencing
Interactive Wardley Mapping Interface
A visual editor to create and annotate Wardley Maps with drag-and-drop capabilities.
- Tools: React, D3.js, Mermaid.js
- Goal: Collaborative decision-making and strategic clarity
AI-augmented Architecture Assistant
A prototype assistant that helps generate architecture diagrams and suggestions based on input requirements.
- Technologies: OpenAI API, Natural Language Processing
- Goal: Bridge the gap between architectural intent and documentation