I am not a software developer. I am a delivery and product person who builds with AI to learn how these tools really work, and to turn ideas into something I can test. One of these projects has had serious attention; the others are early experiments. All the code is open on GitHub, and I would be glad to hear what you think.
CV Tailor
This is the project I have worked on most. It is an AI pipeline that takes a source document and a written brief, checks whether the request is worth acting on, and only then produces tailored, fact-checked Word documents. The interesting part is not the generation. It is the quality bar around it.
What I designed
- Two gates before any output. A twelve-point screening test, then a match score from 0 to 10. Weak cases stop, and borderline ones ask a human.
- Checking the AI without another AI. A rule-based fact checker scans every generated document against fixed factual rules.
- A human stays in the loop. In batch mode, borderline cases are held for review rather than processed automatically.
- Guardrails against invented claims. Every achievement used must be attributed to the right source and reproduced accurately.
- Private by design. Personal details live outside the code, and the tests fail if any of them leak into the repository.
- Tested. An offline test suite runs in about fifteen seconds and costs nothing to run.
Built with: Python, Node.js, Streamlit and the Claude API, with AI assistance throughout.
Early experiments
These two are sketches. They work, but they are early and I have not taken them further yet.
Career Compass
A coaching companion built around the Career Compass Map, a framework for mapping a career across three to six dimensions. AI analysis looks at how well the dimensions fit together, draws a radar chart of the career’s shape, and exports the result as a PDF.
Built with: React, Vite, Flask and the Claude API
Status: early experiment, currently paused
Project Planner
A small tool that turns three simple input files into an Excel plan: a dashboard with a burnup chart, a Gantt-style view by sprint, and a risk log with colour-coded scoring. Think of it as Microsoft Project, super-lite.
Built with: Python and openpyxl
Status: early experiment, currently paused
Before you try them
All three projects are released under the MIT licence, so you are free to use, adapt and share them, as long as you keep the copyright notice. They run on your own computer. There are no hosted demos, and CV Tailor and Career Compass need your own Anthropic API key. They are early and built with AI assistance, so treat them as they are: no warranty and no promise of support. If you find a bug or have an idea, I would like to hear from you on the Contact page.