Full-Stack Software Engineer
|
A few of them, and what broke along the way.
Agricultural management SaaS I work on at OMD do Brasil. I wrote the backend from the ground up in NestJS, Prisma and PostgreSQL, with a layered architecture of domain, use cases and repositories that the team adopted as its standard. I also containerized the application and wrote the deploy pipeline, which is what let us move the platform off Google App Engine.
A resume builder that generates PDFs, with billing in three currencies and routing between Stripe and PIX depending on the country. Interface and content are available in English and Portuguese through next-intl.
I work in the TypeScript ecosystem: NestJS and Node on the API, Next.js and React on the interface, PostgreSQL underneath. I care more about getting the data model right early than about the framework of the month.
I keep up with the newer parts of the stack and ship production code with Next.js, React 19, Tailwind, and Biome. I'll pick up a new tool early when it actually earns its place. Right now I'm open to full-stack roles where I own a feature from the API to the UI.
Sertãozinho, SP, BrazilI build and ship complete SaaS products, from the database schema to the polished UI, using the Next.js App Router, Server Actions, and edge-first deploys.
Built REST APIs with Node.js, NestJS, and Hono. I isolate tenants with Row Level Security in Postgres, applied per request, instead of trusting every query to remember the filter.
Built performance-first frontends with React Server Components, Turbopack, and code-splitting. I profile before optimizing, so the work goes where the time is actually spent.
Anything non-trivial starts as a written spec: the decision, the constraint, and the option I ruled out on purpose. The agent reads it, and so does whoever picks the project up next.
Generated code gets the same pass as a pull request from a stranger. Most of what I throw away looks correct and is quietly wrong in one line.
Business rules and access policies get tests before I trust a change. An agent that can run the suite and read the failure is worth more than one that writes the code faster.
Versioned migrations, decisions written down next to the code, and a project file the agent reads on the way in. That is what lets me pick up a project six weeks later without reconstructing it from memory.
Highlighted items are the ones I use every day.
Working on something, have an idea, or just want to talk about engineering? Reach out through any channel below, or send a message here.