← Back to Work

Headhunter.com.hk: HK Recruitment Directory

tech · 6 weeks
Headhunter.com.hk: HK Recruitment Directory

The problem

Hong Kong has hundreds of recruitment agencies. Finding the right one means Googling, checking LinkedIn, asking friends. No single directory covers them all with real data.

What we did

Built the largest HK recruitment directory: 689 companies, 97 MTR stations, 18 districts, 22 industries. Every company page is a knowledge vault note with wikilinks. Articles compiled from atomic facts, not generated from thin air.

Scope

Data pipeline, content generation, SEO architecture, deployment

Result

730 pages built in 1.15 seconds. Six directory verticals live (headhunters, auditors, renovation, warehousing, water, education). One Astro codebase powers all of them.

The idea was simple: build the directory that Google can’t ignore.

Most directories are databases with a search box. Ours is different. Each company has a knowledge vault note: atomic facts with wikilinks connecting them to districts, MTR stations, industries, and other companies. When you read about a recruiter in Causeway Bay, you see which MTR station is closest, what industries they cover, and what other agencies share the same building.

The content pipeline is the interesting part. We don’t ask AI to “write an article about recruitment in Hong Kong.” That produces generic garbage. Instead, we seed atomic facts from real data (company registries, public filings, location data), then compose articles by connecting those facts. The AI’s job is synthesis, not invention.

One Astro codebase powers six different directory sites. Change the SITE_SLUG environment variable and the same engine builds a completely different vertical: auditors, renovation companies, warehousing, water suppliers, education providers. Zero code changes between verticals. That’s the whole trick.

The SEO architecture is designed for AI search visibility first, traditional search second. Every page answers a specific question. Internal linking follows the knowledge graph structure. FAQs are generated from real data, not made up.

This is the Swarm project. The goal: build the largest network of Hong Kong business directories, then sell premium placement to businesses that want to be found.

Want something like this?

Start a project