Backend-first builder | Systems Development at SENAI PR
I build my backend foundations in public and treat delivery as a system: clear contracts, safe defaults, reproducible workflows, and pragmatic automation. My current focus is Node.js, TypeScript, API architecture, and AI-assisted developer tooling.
- Constraints improve design. Clear boundaries usually produce better systems.
- Explicit contracts beat implicit behavior. Errors, interfaces, and validation should be predictable.
- Reliable delivery depends on repeatable workflows. Checks should be easy to run and hard to skip.
I use AI tools as part of a disciplined engineering loop, not as a replacement for verification. The goal is faster iteration with explicit validation, focused diffs, and lower operational risk.
| Tool | Best for | Guardrails |
|---|---|---|
| Gemini CLI | Research, exploration, alternative approaches, first-pass drafts | Treat outputs as input, verify claims, avoid copy-paste from external sources |
| Claude Code & Ext. | Deep reasoning, complex refactors, and architectural context | Verify logic steps, review before large diffs, enforce strict context boundaries |
| Codex CLI | Code changes, refactors, local checks, focused implementation | Manual-first workflow, explicit commands, small reviewable diffs |
| Antigravity | Structured sessions, scoped agents, targeted task decomposition | Keep a single source of truth, use only what the task needs |
| MCP | Safe connections between tools, services, and project context | Least privilege, explicit boundaries, no unnecessary secret exposure |
- Flows first: Initialize workspace rules and context using explicit scripts (
06-scripts/). - Test-Driven Planning (TDP): Explore the problem space and draft the structural proof plan with Gemini CLI (Plan Mode) for a "Human-in-the-Loop" block.
- Execution: Post-plan, implement iteratively and validate via Codex (CLI/VS Code) or the native IDE.
- Specialists: Call Antigravity or Claude Code when a task benefits from deep reasoning, QA, debugging, or a specialized multi-agent pass.
- Close: Run explicit checks, require adversarial review if risky, and finalize with a concrete summary of tradeoffs.
- Claude Code (docs) - deep reasoning workflows, agentic loop assistance, and complex codebase modifications.
- Codex CLI (npm, overview) - local coding workflows, focused edits, and validation loops.
- OpenAI Codex app (release) - structured coding sessions and project-oriented workflows.
- Gemini CLI (docs, repo) - terminal-first assistance for research, code iteration, and productivity.
- Google Antigravity (announcement, download) - modular agent workflows and scoped sessions.
- Model Context Protocol (MCP) (spec) - safe, scoped tool connectivity across systems.
What I use these for:
- Supabase - schema iteration, RLS patterns, storage workflows, and local-to-prod feedback loops.
- TestSprite - automated test flows, CI-friendly validation, and coverage signals.
- Stripe - payment lifecycle experiments, webhooks, and test-mode checkout flows.
- Sentry - production debugging loops from issue to event to fix plan.
- Clerk - auth quickstarts and modern identity patterns across app frameworks.
| Category | Technologies |
|---|---|
| Languages & Runtime | JavaScript (ES6+), TypeScript, Node.js |
| Frontend | React, Next.js, Tailwind CSS, shadcn/ui |
| Backend | Express.js, REST APIs, Clean Architecture |
| Databases | SQLite, PostgreSQL, Prisma |
| Infra & Delivery | Docker, Linux CLI, Git, GitHub, Vercel |
- Petshop Small Breeds Premium - full-stack practice with admin dashboard, auth flows, booking requests, and Vercel deployment.
- Voice Note AI - Windows-first dictation app with Azure Speech-to-Text, safe text injection, and adaptive suggestions.
- Clean Express API - architecture-first API patterns with validation, consistent errors, and TypeScript-first structure.
- Backend TS Foundations - disciplined Node.js and TypeScript practice with a focus on consistency and delivery basics.
- TradingView Indicator - Pine Script experiments for structured technical analysis and trading automation.
- Ship stronger end-to-end projects with auth, observability, deployment discipline, and secure backend defaults.
- Publish better API baselines with contract clarity, validation, and operational guardrails.
- Keep improving AI-assisted workflows without relaxing verification quality.
A live view of my current local workspace architecture, optimized for agentic development:
projetos/
├── 01-projetos/ # Active projects (frontend & backend apps)
├── 02-pacotes/ # Shared packages (e.g., UI kits, libraries)
├── 03-playground/ # Sandboxes and experiments
├── 04-docs/ # Auxiliary or legacy material
├── 05-arquivo/ # Historical, paused, and temporary material
├── 06-scripts/ # Workspace automation and CLI tooling
├── 07-github/ # Local mirror of external repositories
├── 08-operacional/ # Backups, artifacts, and operational logs
├── docs/ # Canonical documentation hub
├── tests/ # End-to-end and governance testing
├── tasks/ # Active sprint, lessons, and progress tracking
├── plans/ # Transient implementation plans (TDP mode)
├── skills/ # Antigravity/Agent extensions and specific toolsets
├── .agent/ # Agent rules and workflows (Antigravity/Gemini/Claude)
└── mcp-servers/ # Local Model Context Protocol servers
