Lead Engineer  |  Python • Node.js • TypeScript  |  AI/LLM Systems, Full-Stack Architecture & Cloud
🇧🇷 Brazil · 🌎 Remote, working with US companies · 20+ years building production systems

Engineering leader and Principal Full-Stack Engineer who takes post-MVP products and makes them stable, scalable and cost-efficient. I own systems end to end, from architecture and data modeling to CI/CD and production operations, and I partner directly with founders to turn product ideas into technical priorities.

Today I build AI/LLM systems in production: multi-provider LLM gateways with token and cost tracking, prompt caching, RAG with pgvector, LLM-as-judge guardrails and evals, and human-in-the-loop AI workflows, on top of solid full-stack foundations in Python/FastAPI, Node.js/NestJS, React/Next.js, React Native, PostgreSQL, AWS and Cloudflare.

• Sole owner of a production Python/FastAPI AI platform (70 API routes, ~19 background workers, 15 external integrations), taken over from a single-developer MVP and hardened for stability, rate limits and billing correctness.
• Multi-provider LLM gateway serving 32 AI routes across Anthropic and OpenAI, with per-route model and cost tiers, prompt caching, token accounting, outage alerts and automatic recovery; the production vendor switches by configuration, with no code changes.
• Human-in-the-loop AI drafting with multi-step pipelines (classify, retrieve, generate), pgvector retrieval, LLM-as-judge guardrails, bounded self-correction and prompt-injection defenses; the system learns versioned prompt rules from reviewer edits.
• AI-native SDLC with strict quality gates: 3,869 automated tests, golden-set and LLM-judge evals, written review standards for AI-generated code, mandatory Claude + Codex review before every push and deploys with health checks and automatic rollback.
• Open-source author of @bymax-one, a 10-package TypeScript platform for NestJS, React and Next.js (auth, queues, realtime, caching, storage, notifications, logging, config, AI token cost accounting), ~5,400 downloads/month, held to 100% test coverage and 100% mutation score.
• Bymax Fitness, my AI fitness platform live in production: RAG pipeline with OpenAI embeddings and pgvector HNSW, BullMQ workers, an AI agent with memory for explainable workout generation, Stripe subscriptions and WebSocket notifications, architected for 10,000+ concurrent users.
• Bymax Bio, a privacy-first, local-first React Native / Expo health tracker with 515 test suites at 100% coverage and 166 Maestro E2E flows; its Next.js web frontend runs on Cloudflare Workers.
• Full-stack delivery for US product teams: accounting core and tenant payments (ACH, autopay) for a B2B property platform, DocuSign e-signature integration, a zero-downtime API migration to NestJS and real-time crypto products.
• Proven leadership: Tech Lead of a 5-person MVP team, and in industry led the automation of a US$224M rolling mill coordinating 70 people, delivered on time with zero major accidents.

• Lead / Principal Full-Stack Architecture, Python, FastAPI, Node.js, NestJS, TypeScript, React, Next.js, React Native
• AI & LLM systems, multi-provider LLM gateways, token and cost accounting, prompt caching, RAG with pgvector, structured outputs and tool use, LLM-as-judge guardrails and evals, AI agents with memory, human-in-the-loop workflows, MCP
• Event-driven & real-time systems, queues and background workers (BullMQ, Amazon SQS), AWS SNS/EventBridge, Redis Pub/Sub, WebSockets, Server-Sent Events, webhook-driven integrations
• Data & modeling, PostgreSQL design and optimization, pgvector, Redis, Prisma, schema design and API data contracts
• Cloud & DevOps, AWS, Cloudflare, Docker, CI/CD with GitHub Actions, observability with New Relic and OpenTelemetry
• Engineering excellence, AI-native SDLC, review standards for AI-generated code, unit/integration/E2E testing, mutation testing, coverage gates, open-source maintainership
• Technical leadership, leading lean product teams, spec-driven delivery, sprint management with GitHub Issues and Projects, partnering with founders and product

Advanced in English, native in Portuguese and intermediate in Spanish, used to daily standups and async collaboration in distributed, remote-first teams. Let's talk on LinkedIn.