AI-Powered Meta Ads Copilot
Budget: $1,500 – $3,000 USD
I’m building an AI assistant that pulls data and analyze data from the Meta Ads platform. The core of the product is a web app running on Next.js 14 with TypeScript, Supabase (PostgreSQL + Auth), and Vercel for deployment. Tailwind CSS and Radix UI handle the interface, while Recharts renders the visualisations.
The assistant must automatically surface:
• performance trends across campaigns and ad sets,
• shifts in audience behaviour, and
• creative-level observations that explain why specific images, copy, or placements are under- or over-performing.
These insights will show up inside interactive dashboards and downloadable PDF/CSV reports.
Key deliverables
1. Secure data ingestion service that syncs Meta Ads metrics to PostgreSQL on Supabase.
2. Backend functions that call OpenAI, cache results, and expose them via a typed API.
3. Front-end views: overview dashboard, campaign drill-down, and report generator, all styled with Tailwind + Radix and charted through Recharts.
4. Auth flow combining Supabase Auth and Meta OAuth, complete with token refresh logic.
5. Deployment setup on Vercel, including environment variables and build scripts.
Acceptance criteria
• Dashboards and reports load in <2 s using live or cached data.
• Each insight block includes a plain-language summary and a supporting metric or chart.
• Codebase follows Next.js App Router conventions, is fully typed, and passes ESLint/Prettier checks.
If you’ve shipped production apps with Next.js, I’d love to see links or repos demonstrating clean architecture.
The assistant must automatically surface:
• performance trends across campaigns and ad sets,
• shifts in audience behaviour, and
• creative-level observations that explain why specific images, copy, or placements are under- or over-performing.
These insights will show up inside interactive dashboards and downloadable PDF/CSV reports.
Key deliverables
1. Secure data ingestion service that syncs Meta Ads metrics to PostgreSQL on Supabase.
2. Backend functions that call OpenAI, cache results, and expose them via a typed API.
3. Front-end views: overview dashboard, campaign drill-down, and report generator, all styled with Tailwind + Radix and charted through Recharts.
4. Auth flow combining Supabase Auth and Meta OAuth, complete with token refresh logic.
5. Deployment setup on Vercel, including environment variables and build scripts.
Acceptance criteria
• Dashboards and reports load in <2 s using live or cached data.
• Each insight block includes a plain-language summary and a supporting metric or chart.
• Codebase follows Next.js App Router conventions, is fully typed, and passes ESLint/Prettier checks.
If you’ve shipped production apps with Next.js, I’d love to see links or repos demonstrating clean architecture.