# Ulbra Infra — From a person on the box to a repository and a pipeline.

> The internal datacenter platform every other system here deploys onto: one script takes a bare server to ready, applications ship from a repository through CI, and an alert can open its own pull request.

## Overview

The platform underneath every other system in this group. It began as on-premise servers with no automation at all — deployment meant a person on the machine, installing a runtime and starting the application by hand. It is now a provisioning script, a container orchestrator, a reverse proxy, two observability tools with a declared split, and a delivery loop in which an alert can investigate itself and open a pull request.

## Stack

- Docker Swarm
- Traefik
- GitHub Actions
- OpenTelemetry
- SigNoz
- Prometheus
- Grafana
- Metabase

## Highlights

- One script takes a bare server to ready: container runtime, firewall, cluster, overlay networks, proxy and monitoring.
- A new application needs a compose file and a workflow — the routing and the certificate follow from labels.
- Telemetry is opt-in through environment variables; nothing else has to be wired.
- An alert can be investigated automatically and arrive as a pull request for a human to judge.

## Role

Head of Technology — design & implementation · Apr 2026 – Current. Closed-source professional work — the write-up describes it without the source.

---

Source: https://felipealmeida.tech/projects/ulbra-infra