Full-Stack Next.js Developer — Build Large-Scale SEO Website with AI-Generated Content (Vercel, Supabase, ISR)

Job Title
Full-Stack Next.js Developer — Build Large-Scale SEO Website with AI-Generated Content (Vercel, Supabase, ISR)

Project Overview
We are building a large-scale SEO website targeting ~38,000 U.S. cities × 9+ service pages each (? 342,000 total URLs).
Each page will include:

12+ unique local data points (from a provided CSV)

AI-generated, long-form content (Tier A pages: 2,000–4,000 words; Tier B pages: 700–1,200 words)

Our current site is built with ---------- v (Next.js App Router) and deployed to Vercel. The problem is the current build attempts to pre-render all pages, causing timeouts and 404 errors.
We need a scalable, database-driven, ISR-based architecture with automated AI content generation.

This is NOT a small “test project.” This is the full production build.
Payment will be made only on delivery of a complete, live, working deployment meeting the acceptance criteria below.

Core Requirements
You must have verifiable experience in all of the following:

Core Technical Skills
Next.js 14+ (App Router) — dynamic routes, server components, ISR (revalidate, revalidateTag)

Vercel — deployment, environment variables, performance tuning for large-scale dynamic pages

Postgres (Supabase preferred) — schema design, indexing, optimized queries, CSV imports

Node.js — building server-side scripts for bulk imports and processing

Prisma ORM or advanced SQL — whichever you prefer for DB access

JavaScript & TypeScript — production-ready, type-safe code

Performance optimization — preventing cold-start lag, caching strategies

AI & Automation Skills
OpenAI API / Vercel AI SDK — prompt engineering, batching, rate limit handling

Content caching strategies — generate AI content once, store in DB, serve without regenerating

Token cost optimization — modular content generation to reduce API usage

Automated content backfill scripts — pre-generate Tier A pages

SEO-Specific Skills
Large-scale SEO strategy — crawl budget management, E-E-A-T implementation, avoiding duplicate content

XML sitemap generation — chunked sitemaps and sitemap index files

Internal linking strategy — hierarchical linking between state, city, and city+service pages

Structured data ( ---------- ) — LocalBusiness, Service, and FAQ schema

Page performance optimization — Lighthouse/PageSpeed SEO requirements

Nice-to-Have
Experience with deployments exceeding 100k pages on Vercel or similar JAMstack/CDN platforms

Edge caching/CDN optimization

Supabase Edge Functions

Large CSV/ETL workflows

UI integration skills to maintain the ---------- v-generated design while adding dynamic functionality

Scope of Work
1. Data Layer
Import provided CSV into Supabase:

cities (city_slug, name, state)

topics (topic_slug, title)

city_topics (city_slug, topic_slug, data_json) — contains the 12+ datapoints in JSONB

page_content (city_slug, topic_slug, content_html, generated_at, quality_flag)

Add composite indexes for fast lookups

2. Dynamic Routing
/[city]/[service] routes

Query DB for data and content

If content is missing and page is in the generation allowlist, call AI to generate, save to DB, and serve via ISR

Return notFound() only if no matching DB record exists

3. AI Content Generation
Tier A (top 1–3k pages): 2,000–4,000 words

Tier B (long tail): 700–1,200 words

Use prompt templates we provide (injecting city datapoints)

Save generated content permanently (no repeated API calls)

Support manual re-generation via revalidation endpoint

4. Sitemaps
/sitemap.xml ? sitemap index file linking to per-state sitemaps

Each state sitemap lists all city/service URLs for that state (? 45k URLs per sitemap)

Must generate instantly (streamed or cached)

5. Deployment
Deploy to Vercel with proper env vars:

SUPABASE_DB_URL

OPENAI_API_KEY

NEXT_PUBLIC_SITE_URL

Verify all ~342k URLs resolve without build errors or timeouts

Implement 24-hour ISR cache revalidation

Acceptance Criteria
Zero 404s for valid city/service URLs

First-time requests generate AI content + cache via ISR

Subsequent requests load in