AI Product Engineer. I build LLM features and agent systems into products, with TypeScript, React and Next.js. Based in Rio de Janeiro, Brazil, working remotely.
10+ years shipping SaaS, fintech, edtech and e-commerce products. Today I work AI-first: spec-driven development with AI agents, and Claude-powered features designed around structured output, cost control and human approval.
- LLM features with structured output. A sales-funnel analyzer that returns a health score, bottlenecks and recommendations as schema-validated JSON, AI review tools for CVs, LinkedIn profiles and pitches with access gating, rate limits and per-call cost logging, and a notes assistant with hybrid retrieval (lexical + embeddings, pgvector).
- An agent runtime. Tool use, model routing by tier and provider (Anthropic or a local model), bring-your-own-key and credits, Redis-backed rate limiting, and a human-approval gate for irreversible actions.
- LLM output as untrusted input. An LLM turns a prompt into a product blueprint and campaign plan; modules are checked against a catalog, budgets are clamped to a safe range, and a human reviews before anything runs.
- A team of AI agents. C-level leads and specialists with an explicit contract, plus an adversarial reviewer agent.
- AI features with cost controls. With my co-founder at Bizurado, a study-planning SaaS: an AI tutor, essay grading and question generation saved as drafts for human review, behind a global kill switch, daily and monthly spend caps and a per-call cost ledger.
Most of this lives in private repositories (company and client code). What is public here is extracted and sanitized on purpose.
Open to remote AI Product Engineer and AI Engineer roles.
TypeScript · React · Next.js · Node.js · Nest.js · Hono · Claude API · pgvector · Playwright