Senior Full-Stack & AI Engineer · Joinville, Brazil · flabs.tech
10+ years across frontend, backend, and AI engineering — shipping products for companies in the US, Europe, and Brazil. I balance solid architecture with real-world delivery.
| Repository | Description |
|---|---|
| arxiv-manager | AI-powered visual-reasoning image evaluation tool — CRAG (CAG + RAG), self-critique generation pipeline, ChromaDB + sentence-transformers, MCP server, observability, hot-swappable prompts |
| flabs.tech | Next.js 16 portfolio — Vitest, Playwright E2E, visual snapshots, CI, live at flabs.tech |
| agenthood | Society of 16 AI agents — SKILL.md open standard · TS framework · VS Code extension · Academy |
| agenthood-site | Agenthood landing page — agent-agnostic SKILL.md skills, SEO-optimized, Studio playground, live at agenthood.flabs.tech |
| hasheyes | Blockchain explorer — FastAPI, ccxt, LLM-powered AI assistant |
| logroute | FMCSA ELD logbook & route planner — Django + React, live at logroute-app.vercel.app |
- AI agent frameworks: 10-layer TypeScript framework design, multi-agent orchestration (16 agents), ReAct reasoning loops, 5-tier memory, Agentic RAG, Knowledge Graph retrieval, hierarchical chunking, Society self-awareness, MemberOrchestrator — all shipped as agenthood
- Browser-based Studio: Chat playground at agenthood.flabs.tech — Next.js 16 + Tailwind CSS 4 + agenthood runtime, SSE streaming, 6-provider routing with automatic failover, server-side rate limiting, BYOK, Edge middleware, structured logging with secret redaction, vitest, 17 tests
- FMCSA-compliant ELD & route planning: Full-stack Django + React app at logroute-app.vercel.app — HOS engine simulating 6 FMCSA rules (11hr/14hr/30min/10hr/70hr/34hr), real-time OSRM routing, Nominatim geocoding, Leaflet maps, and canvas-based DOT log sheet generation with jsPDF export
- Backend services: Scalable Node.js + TypeScript and Python/Django REST APIs, auth middleware, rate limiting, distributed tracing, PostgreSQL optimization
- Frontend platforms: React 19, Next.js 16, TypeScript, Storybook, performance tuning, Vitest + Playwright testing
- Python apps & voice: Streamlit + FastAPI data dashboards (KPIs, risk panels, alerts, news flows, AI assistant, newsletter with ccxt exchange integration) and Python voice assistants for Android with wake-word, STT/TTS pipelines, and Claude AI integration
- Visual-reasoning image evaluation (arxiv-manager): AI-powered tool generating Challenging/HARDEST Q&A from scientific figures with self-critique loops, consensus drafting, and 28+ handbook validation rules — CRAG architecture (semantic cache + hybrid ChromaDB retrieve + cross-encoder rerank), MCP server (6 tools), hot-swappable prompt registry, structured observability with token/cost tracking, security guardrails (injection detection, PII filter, rate limiting), guardrails with auto-retry, dynamic model selection, and 220 tests — all Python/FastAPI at arxiv-manager
- GenAI & RAG pipelines: Document-grounded assistants, vector store integrations (LanceDB, Chroma, FAISS), Claude AI, semantic chunking strategies, fixed-size and hierarchical chunking
- Quality & CI: Vitest (49+ test files, 544+ tests), Playwright E2E + visual snapshots, 13 CI workflows, conventional commits, PR review automation, eval frameworks, gitleaks secret scanning
Frontend:
Backend & APIs:
Web & Templating:
Python Ecosystem:
AI & Agents:
Testing:
Cloud & DevOps:
- LinkedIn: in/fabiorborges — fastest reply
- Portfolio: flabs.tech — projects, writing, and contact form
- npm: agenthood — latest published package
*Last updated: Jul 25, 2026