Skip to content

Help me improve this site?

I use anonymous statistics to understand which articles and projects interest visitors. No advertising and no marketing profiles, just metrics that help me create better content. You can change your mind at any time.

View Cookie Policy
← Back to Projects
2025·In Progress

AI-Powered Personal Portfolio

Researcher & Implementer

My personal website, built as a lab to experiment with and compare AI agents in web development. I started by testing several tools and ended up consolidating development on Claude, which is now the tool I use to evolve the site.

Overview

Personal project that serves as a lab for AI-assisted development. The goal was not just to build a portfolio, but to understand where each tool adds value, where it falls short and how human oversight remains central, from prototyping to content, accessibility and SEO. Over time the work consolidated on Claude, now the main tool for evolving the site.

Challenge

Develop a responsive, professional portfolio in a short timeframe using AI and no-code/low-code tools while ensuring robustness, scalability and good accessibility and security practices.

Solution

  • Hands-on comparison of AI agents (Lovable, Devin, ChatGPT, Copilot, Claude) in prototyping and development, to understand where each one adds value.
  • Development consolidated on Claude (Claude Code), now the main tool for working on the site.
  • Occasional support from other models for research and content review.
  • Deployed on Vercel, versioned on GitHub, monitored with Google Analytics.

Results

  • Hands-on comparison of several AI agents on the same project, with a clear sense of what each does well and badly
  • Understanding of agent differences: each AI has distinct strengths in code generation, review and content
  • Iterations in minutes rather than hours, significantly accelerating the development cycle
  • Human validation proved essential: AI accelerates but does not replace design and architecture decisions
  • A replicable way of working: one main agent and human validation on every decision

Technologies and Methods

ReactTypeScriptViteTailwind CSSClaude CodeLovableGitHubVercelGoogle AnalyticsLighthouse

Detailed Implementation

01

Planning & Setup

Defined goals, selected AI tools and prepared the collaborative pipeline.

02

Accelerated Prototyping

Used agents and no-code/low-code tools to build prototypes and quickly validate flows.

03

Integrated Development

Built the site with React, Vite and Tailwind, hosted on Vercel, with PT/EN versions, reflections and analytics.

04

Monitoring & Evolution

Configured metrics with Google Analytics and Lighthouse and performed continuous performance and SEO optimisations.

Impact Quantified

10+

AI tools tested and compared in a real context

3 phases

Prototyping, development and refinement with distinct agents

Minutes

Iteration cycles that would conventionally take hours

Lessons Learned

  • Comparing agents side by side teaches more than any benchmark: each has its own strengths and limits
  • Focusing on one main tool pays off more than hopping between several
  • Different AIs bring complementary perspectives and improve content refinement effectiveness
  • Human validation is essential to ensure consistency, compliance and brand alignment
  • The experience strengthened skills in UX/UI, API integration and agile deployment

Next Steps

  • Test autonomous agents for automatic content updates
  • Automate the distribution of reflections (newsletter)
  • Measure the real impact of SEO and performance optimisations on usage data
See Other ProjectsGet in Touch