Building My Portfolio Website with AI-Powered Development

When I set out to build my personal portfolio at fdrechsler.de, I decided to dogfood my own creation: an AI Development Framework that orchestrates multiple specialized agents to automate the entire software development lifecycle.

The Problem with Traditional Development

As a mathematician and software engineer, I have always valued efficiency and systematic approaches. Traditional development workflows, even with modern CI/CD pipelines, still require significant manual effort:

• Writing boilerplate code
• Creating and maintaining tests
• Performing code reviews
• Managing deployment pipelines

"The best way to predict the future is to invent it." - Alan Kay

I wanted something better: a system where AI agents collaborate like a well-coordinated team.

The AI Development Framework

The framework uses a ticket-based workflow where specialized agents collaborate autonomously:

Agent Roles

1. Implementation Agent - Writes production code following project conventions
2. Test Agent - Creates comprehensive test suites and verifies quality
3. Review Agent - Performs code quality checks and suggests improvements

Workflow Stages

Tickets progress through defined statuses:

DRAFT -> READY -> WORKING -> REVIEW -> DONE

During the WORKING phase, agents iterate autonomously until quality standards are met. Human approval is only required at the REVIEW checkpoint before merging.

Technical Stack

The portfolio itself is built with modern web technologies:

• Next.js 16 with the App Router for server-side rendering and static generation
• TypeScript for type safety across the entire codebase
• Tailwind CSS v4 with shadcn/ui components for a clean, responsive design
• Vercel for deployment with automatic preview environments

Code Example: Structured Data for SEO

One interesting aspect was implementing structured data for both traditional search engines and AI-powered search:

export const personSchema: WithContext<Person> = {
  "@context": "https://schema.org",
  "@type": "Person",
  name: "Dr. Florian Drechsler",
  jobTitle: "Software Engineer & AI Developer",
  knowsAbout: [
    "Machine Learning",
    "Artificial Intelligence",
    "Mathematics",
    "Software Engineering",
  ],
}

This helps search engines and AI assistants understand who I am and what I do, making my profile more discoverable.

Key Learnings

Building with AI agents taught me several valuable lessons:

What Works Well

• Repetitive tasks like generating boilerplate, tests, and documentation
• Consistent code style across the entire project
• Rapid iteration on features with automated quality checks
• Comprehensive test coverage that humans often skip under time pressure

What Requires Human Judgment

• Architecture decisions that affect long-term maintainability
• UX design choices that require empathy and user understanding
• Security considerations where mistakes have real consequences
• Creative content that needs authentic voice and perspective

Results and Metrics

The AI Development Framework has proven effective in practice:

| Metric | Value | |--------|-------| | Tickets Completed | 41+ | | AI Sessions | 16+ | | Average Time per Feature | Significantly reduced | | Test Coverage | Consistently high |

What is Next

I am continuously improving the framework and plan to:

• Add more specialized agents for specific domains
• Improve the feedback loop between agents
• Open-source components that could benefit the community
• Write more about specific technical challenges and solutions

If you are interested in AI-powered development or want to discuss the framework, feel free to reach out via the contact form or connect with me on LinkedIn.

This blog post was written by a human (me!) reflecting on the experience of building with AI agents. The irony of writing about AI development without using AI to write the post is not lost on me.