Building the HKBP church website was not just a coding exercise. It was a real project for a real community, and at the same time a practical test of my AI-assisted development workflow.
Project Context
I built this as a personal contribution to my church community. The goal was simple:
- make important information easier to find
- centralize announcements and event schedules
- provide a maintainable foundation for future features
This made the project useful as both a nonprofit contribution and a portfolio-grade case study.
Why I Used AI Agents
As a solo engineer, speed matters, but quality still has to hold. I used AI agents to reduce repetitive work while keeping architecture and final decisions under manual review.
AI was most helpful for:
- Planning acceleration
Drafting requirement breakdowns, implementation sequences, and task lists. - UI scaffolding
Rapidly iterating component structures and visual patterns before refinement. - Backend boilerplate
Generating API scaffolds and baseline CRUD flows to shorten startup time. - Debugging support
Faster issue triage and validation of alternative fixes.
Delivery Approach
My workflow for this project:
- Define requirements and scope for MVP.
- Draft PRD and implementation milestones.
- Build initial frontend and backend slices with AI assistance.
- Review and refactor for maintainability.
- Verify behavior, then prepare deployment configuration.
The key principle: AI accelerated execution, but engineering judgment controlled direction.
Tech Stack
- Frontend: React + Tailwind CSS
- Backend: Go (REST API)
- Database: MySQL
- File Storage: Cloudflare R2
- Deployment: Docker + reverse proxy workflow
What Worked Well
- Faster iteration on page and component structure
- Reduced boilerplate time on both frontend and backend
- Better momentum during debugging and implementation handoff
Constraints and Tradeoffs
AI output still needs careful review for:
- architecture consistency
- code readability and maintainability
- security and data handling
- production readiness
In short, AI improved throughput, but not accountability. Final quality still depends on disciplined engineering review.
Closing
This project strengthened my approach to building production-focused software with AI as a force multiplier. In upcoming posts, I’ll break down specific technical decisions for architecture, content workflows, and deployment.