1// Independent product · 20262export constavericode = {3role: "Creator & developer",4stack: ["Next.js", "TypeScript", "PostgreSQL", "Prisma", "Auth.js"],5decisions: [6 { title: "Starting with the team", why: "Workspaces group repositories and expected contributors, giving project leaders a reference for following their team." },7 { title: "Events and scheduled updates", why: "Webhooks receive GitHub events; scheduled synchronizations update the information queried by the application." },8 { title: "Workspace-scoped persistence", why: "Stored data follows the workspace structure, connecting repository information to each team's overview." },9 ],10caseStudy: "/en/projects/avericode/",11};
1// Freelance · Nov 2025 – Mar 20262export constassistenteIa = {3role: "Freelance developer",4stack: ["Node.js", "TypeScript", "Express.js", "PostgreSQL", "Supabase"],5decisions: [6 { title: "Explicit command execution", why: "Some actions were routed through commands and menus executed in code rather than sending every request to the model." },7 { title: "Context for conversations", why: "I improved AI instructions to guide the assistant's behavior within messaging flows." },8 { title: "Validation before use", why: "I checked changes locally and then with real users, verifying integration flows and product behavior." },9 ],10caseStudy: "/en/projects/assistente-ia/",11};
A published bill-splitting product using AI to read receipts.
Technical decisions
A receipt as input. AI reading is part of identifying bill information to help participants split costs.
From interface to infrastructure. My implementation covered the mobile frontend, part of the backend and deployment of the services forming the infrastructure.
Planning the next features. I planned advertising and in-app purchases and assigned implementation tasks to another team.
1// Inteli Júnior · 20252export constporPartes = {3role: "Mobile frontend, backend and infrastructure",4stack: ["Vercel", "Supabase", "OpenAI", "AWS"],5decisions: [6 { title: "A receipt as input", why: "AI reading is part of identifying bill information to help participants split costs." },7 { title: "From interface to infrastructure", why: "My implementation covered the mobile frontend, part of the backend and deployment of the services forming the infrastructure." },8 { title: "Planning the next features", why: "I planned advertising and in-app purchases and assigned implementation tasks to another team." },9 ],10caseStudy: "/en/projects/por-partes/",11};
A distributed backend to collect, normalize and process events from different platforms.
Technical decisions
Partner-provided data. Loading the dump into RDS made the dataset supplied by Globo available to the application.
Windowed replay. Periodic consumption in 15-minute windows simulated a time sequence for processing.
Application and infrastructure. Flask provided the API; Gunicorn and Nginx formed part of the backend runtime setup. My work included this foundation and part of the AWS services.
1// Inteli + Grupo Globo · August — September 20262export constgloboTelemetria = {3role: "AWS infrastructure and Flask backend",4stack: ["Python", "Flask", "Gunicorn", "Nginx", "AWS EC2", "RDS", "S3", "CloudWatch"],5decisions: [6 { title: "Partner-provided data", why: "Loading the dump into RDS made the dataset supplied by Globo available to the application." },7 { title: "Windowed replay", why: "Periodic consumption in 15-minute windows simulated a time sequence for processing." },8 { title: "Application and infrastructure", why: "Flask provided the API; Gunicorn and Nginx formed part of the backend runtime setup. My work included this foundation and part of the AWS services." },9 ],10caseStudy: "/en/projects/globo-telemetria/",11};
1// Hackathon · 1st place ICP WCHL · 20252export constkaisen = {3role: "Developer in a team",4stack: ["TypeScript", "Next.js", "NestJS", "Prisma", "PostgreSQL", "Gemini"],5decisions: [6 { title: "A goal as input", why: "Generation starts with the learning goal and produces sections and content for the path." },7 { title: "Questions with context", why: "Questions sent to the model combine the user's message with the learning-path context." },8 { title: "Gemini in the prototype", why: "The team chose Gemini considering free-tier limits. Specialized minigame calls returned JSON data." },9 ],10caseStudy: "/en/projects/kaisen/",11};
A web/mobile application and graph-based algorithm to support resource allocation for forest fire response.
Technical decisions
Inputs for recommendations. The solution considers distance, risk and availability when allocating fire-response resources.
The backend as the connection. My work connected the interface to the algorithm through the Java backend, using Spring Boot in the application.
Complementary responsibilities. The team developed the application and graph logic; my responsibility was the backend and its integrations, not the algorithm itself.
1// Inteli + Suzano · Feb – Apr 20262export constsuzano = {3role: "Java backend and integrations",4stack: ["Java", "Spring Boot", "TypeScript", "Next.js", "Supabase", "Docker"],5decisions: [6 { title: "Inputs for recommendations", why: "The solution considers distance, risk and availability when allocating fire-response resources." },7 { title: "The backend as the connection", why: "My work connected the interface to the algorithm through the Java backend, using Spring Boot in the application." },8 { title: "Complementary responsibilities", why: "The team developed the application and graph logic; my responsibility was the backend and its integrations, not the algorithm itself." },9 ],10caseStudy: "/en/projects/suzano/",11};