My name is Marcelo. I'm a computer engineer from ITA, a former Google engineer, full-stack developer with 15+ years of experience, and an entrepreneur.
I spent 4.5 years at Google1 before quitting to explore something new: building developer tools, micro-SaaS, and AI-native applications.
These days I work with an "Engineer in the Loop"2 approach: I use LLMs and AI agents like Roo Code, GitHub Copilot, and others as teammates to ship high-quality code faster, while keeping architectural oversight and maintaining the code quality standards I learned at Google.
Developer Tools:
- vscode-improved: A VS Code extension bundling quality-of-life improvements. Built to improve my own workflow while learning the Extensions API.
- ghostkeys: A keyboard remapper in Rust. Started with Kiro as a spec-driven project; now migrating to Roo Code to validate the GitHub Spec-Kit workflow.
- devmagic: A devcontainer setup that "just works", built in public to help developers avoid configuring the same things over and over again.
- brave-byom-compat: Compatibility layer between AI models and Brave's Leo AI.
AI-Powered Stuff:
- concept-compass (Hackathon): An infinite mind-map generator using OpenAI's OSS models. Built it to test and learn Kiro (an AI-powered, spec-driven coding assistant).
- Viralei (Hackathon): Turns legislative bills into engaging videos using AI. Built and deployed the MVP in 4 days, with Next.js. Still working on it.
- Solo Founder AI Engine (In Progress): An integrated system orchestrating multiple AI tools: VS Code with Roo Code, Copilot, and Codex, plus n8n and Dify for workflow automation. Being designed to help solo founders ship like a team.
Products & Companies:
- Futurin (Founder): My software lab where I build and deploy micro-SaaS products using Next.js, TypeScript, Tailwind CSS, and shadcn/ui.
- Adapta AI Challenge: Selected as one of ~80 finalists (from 2,000+ applicants) for Brazil's largest AI Hackathon. Built an OOH analytics tool in ~24h, placed Top 3, and presented to ~7,000 people at the Adapta Summit main stage.
I wrote my first line of code in 2010, in C. Then I learned Java, Python, JavaScript, and more. These days I work mostly with TypeScript, JavaScript, React, Next.js, Node.js, and Tailwind CSS, along with plenty of other tech when they are called for.
At Google I worked with C++, Python, Java, and Bazel (build systems). I also have experience with VS Code Extensions API, Chrome Extensions API, Docker/Podman, and GitHub Actions.
For AI work, I recently adopted Roo Code and it seems like it's going to be central to how I build: I'm using it for custom modes now and will try soon for spec-driven development. I also use GitHub Copilot (in VS Code and github.com), the Codex extension, Jules (jules.google.com), and Kiro. I've also used Claude Code, Gemini CLI, and Zed (zed.dev). I use Anthropic API and OpenRouter as model providers. Lots of prompt engineering.
From Google:
- The value of code reviews, testing, documentation, and treating maintainability as a feature
- How to work with strict 6-week release cycles on a massive C++ codebase (Chromium)
- How to build globally scalable data pipelines (Java/Flume)
- That good practices are fundamental when creating great software
What I believe now:
- Solid foundations matter: Algorithms, system design, and architecture knowledge are essential to use AI tools effectively. How will I know the AI gave me a good solution if I don't understand the bounds?
- AI as a teammate: LLMs are powerful productivity multipliers, but human judgment on design and architecture is still essential.
- Ship fast, maintain well: High velocity with strong architectural oversight, not one or the other
- Build in public: I share my work openly, contributions welcome on all my repos. Developer tools improve fastest when shaped by the community using them.
My path back to building: After Google, I spent time in digital marketing, learning user psychology, sales copy, and how to take products to market. That experience sharpened how I think about user-facing products. Now I'm back to what I do best: building. AI gives me the leverage to ship like a team, without leaving the keyboard.
Feel free to:
- Check some projects and tech that I like
- Read about learnings I had so far
- Read about me (includes the full story of why I quit Google)
- Read some of my content: English | Brazilian Portuguese
- Contact me at: reachme [at] marcelo [dot] land
- Visit my portfolio or LinkedIn
See my pinned repos belowโthey're full of useful stuff!
Education: B.S. Computer Engineering, ITA (Brazil's top engineering school) | Exchange year at University of Twente & TU/e (Netherlands)
Code is Apache-2.0 (LICENSE). Text & images are CC BY-NC 4.0 (LICENSE-CONTENT). Third-party assets belong to their owners.
Footnotes
-
Working mostly with C++, Python and JavaScript. โฉ
-
I discovered this term in Roo Code's job posting. It matched how I already think about AI-assisted development. โฉ




