Implement AI
The production backend for an AI platform: async FastAPI, background jobs and Stripe billing, shipped with Docker and GitHub Actions.
- Role
- Software Engineer, Techanzy
- Stack
- FastAPI · PostgreSQL · Celery
- Live
- implementai.io ↗
Hover or tap a box to see what it does, or press play.
The problem
Implement AI needed a backend that could serve organisations, teams and billing reliably, while keeping slow work out of the request cycle.
What I built
- Async FastAPI on async SQLAlchemy and PostgreSQL, with every schema change as an Alembic migration
- JWT auth and role-based access checked on every route
- REST APIs across organisations, Stripe billing webhooks, documents, CRM/teams and scheduling
- Background jobs on Celery with RabbitMQ and Redis, so slow work never blocks a response
- Docker images deployed through GitHub Actions
How it works
- A request leaves the Next.js client, carrying a JWT.
- FastAPI checks the token and the caller's role before doing anything.
- Reads and writes go to PostgreSQL.
- Work that shouldn't block the response is queued on RabbitMQ.
- A Celery worker picks the job up in the background…
- …and stores its result in Redis.
- Billing runs through Stripe, reported back by its webhooks.
- All of it ships as Docker images through GitHub Actions.
Results
- Serves 500+ API requests a day
- 15+ REST endpoints across the product's areas