name: Ary Rabelo
role: Senior Full-Stack Software Engineer
experience: 20+ years (Ruby on Rails since 2012)
location: Brazil (remote; contracting with US teams since 2019)
status: open to senior fullstack and AI / agent-tooling roles
most_recently: Senior Fullstack Engineer at Canix (2024-2026)
what_i_do_best:
- Keep fast-changing systems reliable (parsers, APIs, flaky CI to trusted green)
- Own features end to end on Rails + GraphQL + React
- Structured-data pipelines: messy input in, clean JSON out
- AI in production: agents, MCP-shaped tooling, human-in-the-loop gates
interests:
- AI and LLMs in production
- Developer tooling and automation
- Open source and community- serpapi-mcp PR: upstream contribution to SerpApi's official MCP server. Added an offline pytest suite built from the real library types and a tests CI workflow, and fixed a no-API-key crash the tests surfaced (issue #35 / PR #36).
- serpapi-agent-toolkit: context-cheap SerpApi search for AI agents. A zero-dependency Bun + TypeScript CLI plus a Claude Code skill, 37 fully-offline tests (MIT).
- I measured MCP vs a CLI for agent search (dev.to): same query, same library, the MCP carried about 17x more tokens per call than a CLI. Reproducible method and honest caveats.
- claude-code-harness: merged PRs to a Go agent-orchestration harness (guardrail i18n; a worktree-hook fix with 17 Go tests plus shell parity).
Languages
Backend & Frameworks
Frontend & Mobile
AI & Automation
Cloud & DevOps
Most production work is in private repos. Recent open source and writing are above.