Building this site as an agentic coding project

Why this site exists

This website serves two purposes. The first is straightforward: present me as a data and AI professional, with applied experience across analytics, road safety, transport, financial services, and technical communication.

The second is less obvious from the outside: the site itself is a practical exercise in agentic coding, using AI tools to help plan, build, and refine the project under a structured, human-reviewed process.

A personal website can easily become one of two things: a static résumé or a personal blog. I wanted neither. Instead, this project is a professional portfolio and credibility asset, built in public, using a workflow I can also describe and stand behind professionally.

What “agentic coding” means in practice

Agentic coding, in this project, does not mean handing an AI tool the whole task and accepting whatever comes back. It means something closer to structured delegation:

• Define the product first: audience, positioning, content boundaries, and tone, written down before any code exists.
• Break the work into small GitHub issues, each with clear acceptance criteria.
• Ask the AI tool to plan an approach before editing anything.
• Keep each change narrowly scoped to one issue.
• Review the resulting diff line by line, rather than assuming it is correct.
• Validate locally, including running the build, before considering anything done.
• Use a pull request as a checkpoint, with a second AI tool reviewing the diff against the issue’s acceptance criteria before it merges.

That loop, plan, scope, implement, review, validate, merge, is where the agentic workflow matters most. The AI tools involved (I’ve used Claude Code for planning and implementation, and Codex as a second-pass reviewer) do meaningful work inside each step, but the scoping, review, and final judgement stay with me.

What I’m learning

Beyond the finished pages, the project is a deliberate way to practise:

• Managing AI-assisted development without giving up engineering judgement
• Structuring a content-driven website with clear information architecture
• Deploying and iterating on a public professional project
• Explaining data and AI work clearly to both technical and non-technical audiences

Why this matters professionally

A portfolio that only shows finished screenshots doesn’t say much about how someone actually works. This project is intended to show the parts that don’t usually make it into a case study: how I scope ambiguous work, how carefully I review output I didn’t write by hand, and how I communicate technical decisions to people who weren’t in the room for them.

Those are the same skills the rest of this site is trying to demonstrate through road safety, transport, and financial services work, just applied here to the process of building the site itself.

Where this stands

This site is still being built, one issue at a time. Some pages are further along than others, and this post itself was drafted and reviewed through the same process it describes: planned, scoped, implemented, and checked against its acceptance criteria before publishing.