← All work
After hours / Backend pipeline
Deal automation
Finds a deal, checks it with AI, posts it to Telegram.
A five-service pipeline on Docker Compose, with Prometheus and Grafana tracking what each service processes and which ones fail.
5pipeline services
10containers in the stack
Geminichecks each deal
Role
I built it. MVP done, analytics in progress.
Stack
Node.js, Docker Compose, PostgreSQL, Redis, Gemini, Prometheus, Grafana
Links
The problem
Finding, checking and formatting deals by hand is repetitive, and ad hoc scripts do not share a queue or handle duplicates.
What I built
- Separate discovery, analysis and Telegram publishing services on Docker Compose.
- Gemini checks each deal before anything gets posted.
- PostgreSQL for persistence and Redis for queues.
- Prometheus and Grafana dashboards that track how much each service processes and spot the ones failing.
Why it is split up
Each worker scales on its own over a shared database and queue, rather than one monolithic job.
MVP plus WhatsApp ingestion. Analytics in progress.