Cloud & automation
Radar de Produção
A newsletter that only ships when every claim has a source.
A serverless AWS pipeline that collects, classifies and drafts a production-engineering newsletter — and blocks the issue if any quality gate fails.
- Status
- Private
- Year
- 2026
- Category
- Cloud & automation
About the project
Radar de Produção (“Production Radar”) follows 96 official sources — AWS, Kubernetes, CNCF, OpenTelemetry, Prometheus, Grafana, HashiCorp, OpenAI, Anthropic and engineering blogs — across five tracks: cloud, Kubernetes and cloud native, observability and SRE, security and deprecations, and AI for operations.
The pipeline is fail-closed. A language model classifies and drafts with structured JSON output, but an issue is only published when every check passes: each claim needs an evidence excerpt and a source, numbers must appear in the evidence, links are checked, duplicates are blocked and the source failure rate cannot exceed 20%. Subscriptions use double opt-in, token-based unsubscribe and SES bounce and complaint handling.
Highlights
- Ingestion every 6 hours and a weekly issue orchestrated by Step Functions, with retry and catch at every step.
- Quality gates: evidence for every claim, numbers checked against the source, valid links, deduplication and a minimum item count.
- Python 3.12 Lambda functions with pyright in strict mode, ruff and an automated test suite.
- Delivery through SES v2 with a configuration set, custom MAIL FROM and bounce and complaint events via SNS.
- State in DynamoDB, artifacts in versioned, encrypted S3 and SQS queues with a dead-letter queue.
- Infrastructure fully in Terraform: CloudWatch alarms, GitHub OIDC with no static keys and an SCP restricting regions.
Projects
More projects
- Active
Developer environment · 2025
Dotfiles
A reproducible Linux environment, from shell to Nix.
Configs and scripts that rebuild my development environment on Ubuntu, Pop!_OS and Fedora — with Zsh, multi-client Git and a portable Nix and Home Manager profile.
- Shell
- Zsh
- Python
- Nix
- +2
- Active
Web · 2025
This portfolio
File-based content, server-rendered, in Portuguese and English.
Next.js 16 with React Server Components, JSON and Markdown content, Portuguese and English routes, dynamic Open Graph images and a PDF resume generated from the same data.
- Next.js
- React
- TypeScript
- Tailwind CSS
- +2
- Alpha
AI tooling · 2026
FalaTrace
Remember what was decided. Check where it was said.
Records and transcribes calls on Linux, links every summary point to the exact moment in the recording and gives AI agents context with a verifiable source.
- TypeScript
- Bun
- Qt/QML
- Python
- +4
Contact
Let's talk?
Got a platform, cloud or observability challenge — or want to chat about one of the projects? Reach me on WhatsApp or by email.