UK Visa Sponsor Search

The problem#

The UK Home Office publishes the Register of Licensed Sponsors, the authoritative list of employers allowed to sponsor work visas, as a flat CSV of more than 125,000 rows. It is refreshed regularly but is effectively unusable in raw form: you cannot search it, filter it, or deep-link to a company. UK Visa Sponsor Search rebuilds that dataset as a fast, SEO-first product.

Architecture#

A single Next.js 16 App Router codebase (React 19, TypeScript in strict mode) that colocates the API routes, the rendered pages, and the data-ingestion pipeline.

src/
  app/
    api/            search, autocomplete, check, export, stats, sponsors, admin/ingest
    cities/[slug]/  dynamic city pages
    counties/[slug]/
    routes/[slug]/  visa-route pages
    sponsors/[slug]/ individual sponsor pages
    sitemap.ts / robots.ts
  components/        layout, search, seo, sponsors, ui (shadcn)
  hooks/            useDebounce, useBookmarks
  lib/
    db/             Mongoose connection + Sponsor model
    ingestion/      fetch -> parse -> transform -> load
  types/sponsor.ts

Data ingestion pipeline#

The register stays current through a four-stage pipeline in lib/ingestion, triggered by npm run ingest or a protected POST /api/admin/ingest route (guarded by an ADMIN_API_KEY) and scheduled with GitHub Actions:

1. Fetch the official GOV.UK CSV.
2. Parse it into structured records.
3. Transform: normalise locations, deduplicate, and generate URL-safe slugs for every sponsor, city, county, and visa route.
4. Load: upsert into MongoDB Atlas.

Search and data model#

Sponsor documents (name, visa routes such as Skilled Worker and Health & Care, city, county, sponsor type, rating) live in MongoDB Atlas via Mongoose 9, with text indexes on the searchable fields.

• GET /api/search runs full-text search with filters for visa route, city, county, and sponsor type.
• GET /api/autocomplete returns type-ahead suggestions, debounced on the client with a useDebounce hook to keep request volume down.
• GET /api/check verifies in a single call whether a given company sponsors visas.
• GET /api/export streams the current filtered result set as CSV.
• GET /api/stats aggregates register-wide totals and breakdowns.

Rendering and SEO#

City, county, and route pages are statically generated, while individual sponsor pages use dynamic [slug] routes, all server-rendered for crawlability. SEO is treated as a first-class feature: sitemap.ts emits entries for every sponsor and category page, JSON-LD BreadcrumbList data and breadcrumb navigation ship on each page, and Open Graph metadata is composed in the root layout.

Stack and deployment#

Next.js 16, React 19, TypeScript (strict), MongoDB Atlas with Mongoose 9, Tailwind CSS v4 with shadcn/ui, and Framer Motion. The UI adapts from tables on desktop to cards on mobile. Deployed on Railway with a GitHub Actions CI/CD workflow; configuration lives in railway.toml, and secrets (MONGODB_URI, ADMIN_API_KEY, NEXT_PUBLIC_APP_URL) are injected per environment.