ProjectsJanuary 5, 2026

Building This Portfolio: AI-Assisted Development in Practice

Overview

This portfolio is itself a case study. It demonstrates something I believe is increasingly true: the boundary between "technical" and "non-technical" work is dissolving. Professionals who combine systems thinking, precise communication, and AI tooling can now build production-grade software without writing code from scratch. The output is indistinguishable from developer-built work. This site is evidence of that.

The Constraint

I needed a portfolio that could compete visually and technically with sites built by engineers. The standard options (Wix, Squarespace, Webflow) trade real flexibility for ease of use. I wanted neither the limitations nor the generic aesthetic that comes with them. The constraint: no deep coding background, high visual and technical standards, and zero tolerance for drag-and-drop ceilings. The solution: reframe what "building" means.

The Methodology: AI as a Development Team

Rather than learning React and TypeScript from scratch, I structured the process as an AI-assisted product workflow, positioning myself as the product manager, architect, and QA lead, and treating AI agents as the execution layer.

1. Foundation Selection

I started with Magic Portfolio, a production-grade open-source template built on Next.js. Beginning with a solid architectural base meant AI agents were modifying existing, well-structured patterns rather than generating from nothing. This dramatically reduced error rates and kept the codebase coherent.

2. Prompt Engineering as Specification Writing

Every modification was driven by precise, context-rich prompts. Vague instructions produce vague code. I learned to communicate requirements the way a product manager communicates to engineers: specific inputs, defined outputs, and clear success criteria. The quality of the prompt determined the quality of the result.

3. Iterative Debugging

When errors occurred, I fed complete error logs back to the AI with full context about the intended behavior. This created a tight debugging loop that resolved issues faster than traditional research, and in the process, taught me the underlying logic of the system I was building.

4. Continuous Deployment

The site runs on Vercel, connected directly to a GitHub repository. Every approved change deploys automatically. No manual uploads, no build commands. Updates go live in under two minutes. The domain erickmahecha.com is custom-purchased and configured end-to-end.

What Was Built

This isn't a template with swapped text and placeholder images. The build includes custom routing and content architecture, modified component layouts across multiple sections, gallery logic that handles both images and embedded YouTube video with dynamic aspect ratio, SEO metadata and Open Graph image generation, password-protected routes for selective content visibility, and integrated newsletter infrastructure. The stack:

Next.jsTypeScriptOnce UIMDXVercelGitHub

AI tools used in development:

GeminiClaude

The Broader Point

Syntax is increasingly abstracted away. What becomes more valuable in its place is the ability to define problems precisely, design systems logically, and communicate requirements clearly enough for an AI to execute them reliably. That is exactly what this project required. That's not a lesser form of technical work. It's the direction technical work is heading.

Related projects

Quick Metal Shop & Dakoma Roofing: AI-Powered Ad Creatives

3.92xROAS

How a lean, one-person production pipeline, built entirely on AI avatars, cloned voices, and rapid editing, replaced traditional video shoots and drove measurable results across two brands, two industries, and two ad objectives.

Read case study

Quick Metal Shop: The Brand Comes to Life

How a brand manual built from the ground up became one of the standout booths at ArquiExpo 2026, and a corporate video carried by a fully original soundtrack, composed with AI.

Read case study

Quick Metal Shop: ArquiExpo 2026 Interactive Gallery Application

48hdev to deploy

Custom-built, vertically-oriented interactive touchscreen web application showcasing 285 architectural panel variants, designed, developed, and deployed in 48 hours for ArquiExpo 2026.

Read case study

Quick Metal Shop: Building a Marketing Department From the Ground Up

Establishing a complete full-stack marketing department and digital infrastructure from scratch, including brand manual, CRM automation, paid media, and AI content engines.

Read case study

Dakoma Roofing: Freelance Rebranding & Digital Automation

Complete brand overhaul, digital channel modernization, AI-powered SEO automation, and premium ad creative production for a B2C roofing and construction services provider.

Read case study

ATV Riders: WhatsApp Lead Automation with Kommo CRM

200+leads/day

End-to-end WhatsApp automation system built for a high-volume ATV dealership, filtering 200+ daily Meta leads through an intelligent chatbot, routing only sales-ready prospects to the commercial team.

Read case study

LeadBot AI: WhatsApp Automation for Lead Qualification

85%conversations automated

AI-powered WhatsApp chatbot that automated 85% of customer conversations, delivering qualified B2B leads to sales 24/7 without adding headcount.

Read case study