Amador, from prototype to live marketplace

At a glance

Amador is a two-sided marketplace where brands find creators for user-generated content. I joined as sole lead engineer when the product was an inherited prototype, and I took it live in about four months.

The situation

Amador was selected for the Antler Australia pre-seed residency. When I joined, the product was an inherited prototype.

I work with the founder on technical direction and product.

What I did

I took the prototype to a live production platform.
I built campaigns, brand surveys and polls.
I designed the points ledger and a tiered subscription model.

How I used AI

In the product

I built an onboarding flow that uses the OpenAI API to write a brand identity and community kit.
Each brand has a token balance that meters its AI use.

In my engineering work

I use Claude Code. Each change goes through a spike, a spec and a review before the build.
Review agents check security, database, Stripe and UX changes.
Scheduled agents triage dependency updates, fix issues, review pull requests and reduce type errors.

Keeping it safe

Each agent has a written scope and read-only access to production.
Nothing pushes to main. A person approves every merge.

The result

More than 140 brands and creators use the platform.
The founder can release changes safely, through a guarded workflow.
Amador was accepted into the Vercel for Startups programme.

Technical detail

Payments and data

The subscription model has campaign-slot limits, scheduled plan changes and Stripe webhook reconciliation.
I made the Postgres functions that move money atomic and idempotent.

Security

I added Row Level Security audits, and rate limiting at a Cloudflare edge Worker.

Delivery

I run the product with trunk-based development, pull request gates, CI and a typecheck ratchet.

Stack

Next.js, React and TypeScript on Vercel. Supabase for Postgres and authentication. Stripe for payments and subscriptions. Cloudflare for the edge Worker. OpenAI for the onboarding flow.