Technical reference for a group of products
I set the architecture, build the difficult parts and write down why, in decision records that every project keeps. The work covers WhatsApp automation with n8n, local AI models, and web and mobile apps.
users.ts
import { db } from "./db"const result = await db.query(const result: Result<{ id: number; email: string; created_at: Date;}[], QueryError> `SELECT id, email, created_at FROM users WHERE active = true`)Recent work
Cut the audit screens from five requests to one
I was the only developer on the core product, so the architecture, data security and production were mine. The audit views needed deeply nested data and made five REST calls to get it. I added a GraphQL layer to the ASP.NET Core backend with Hot Chocolate, so each view now loads in a single request. I also set up Azure AD sign-in with access control, and rewrote the time-record rules so audit data matches to the millisecond.
C#, ASP.NET Core, GraphQL, Azure AD, CI/CD
Shipped the mobile apps for a SaaS with 10,000+ users
I set up the React Native and Expo architecture and a shared TypeScript component library that 3+ apps are built on. Data moved from REST to Apollo GraphQL with a normalized cache, so the apps keep working offline. Releases to iOS and Android go out through GitHub Actions and EAS. I also mentored two junior developers.
React Native, Expo, Apollo, Firebase
Moved a Vue 2 dashboard to Vue 3 while 500+ businesses kept using it
I planned and ran the migration to Vue 3.5 and the Composition API without taking production down. Scattered REST calls became typed Apollo queries generated from the schema, and the heaviest views moved onto Web Workers.
Vue 3, TypeScript, GraphQL Code Generator, Web Workers
Made 50,000-row spreadsheets usable in the browser
A SaaS for 1,000+ enterprise clients across APAC, with real-time views over 100,000+ records. I moved spreadsheet rendering to HTML Canvas and pushed the heavy computation to Web Workers. I also built the companion React Native app on the same GraphQL API.
Nuxt, Canvas, GraphQL, React Native, .NET reporting
Open source
OwlSQL
You write normal SQL. OwlSQL parses it inside the type system, infers the result from your schema, and returns a Result instead of throwing. There's no ORM, no query builder and no codegen step.
I wrote and maintain all of it: the type-level parser, schema inference, the drivers, docs, CI and the npm releases.
OwlSQL on GitHubpieces-to-agents
Turns Pieces long-term memory into AGENTS.md files scoped to a single repository, so coding agents start with real project context. It runs locally and redacts before anything gets written.
Source on GitHubDepFence (in progress)
A self-hosted scanner for the npm supply chain. It statically analyzes lifecycle scripts and Phantom Gyp attack vectors before a package gets the chance to run.
Experience
- Sep 2026 – presentSenior software architect at ATRIA Corp, remote
- Feb 2026 – Jun 2026Software engineer at ohubdev, Balneário Camboriú
- Jul 2025 – Feb 2026Mobile engineer at H2K Tecnologia, Brusque
- Feb 2025 – Jan 2026Frontend developer at Beeasy Digital, remote
- Aug 2023 – Apr 2025Software developer at DJIN Tech Solutions, remote with an international team
- Feb 2022 – Aug 2023Support technician and test analyst at Bravuserp, an ERP and POS used by 200+ businesses
- Jan 2019 – Feb 2022Freelance developer. Five apps on the App Store and Google Play, with Stripe, Pagar.me, WhatsApp and Correios integrations
On the side I run Veltro Sistemas, a small studio that builds software for local businesses. I also have a background in penetration testing, which mostly shows up as a habit of reading code the way an attacker would.
Tools
Most days: TypeScript, Vue and Nuxt, React, React Native, Node.js, C# and ASP.NET Core, GraphQL, PostgreSQL.
Infrastructure: Docker, AWS, Azure DevOps, GitHub Actions, Supabase.
Also used: Rust, Python, PHP, Flutter, n8n, local LLMs.
Languages: Portuguese (native), English (professional), Spanish (basic).