Hey 👋, I'm an AI Engineer
© 2026
001 — The work
Founder & Engineer
AI-powered customer service for restaurants and bars. WhatsApp conversations handled end-to-end by Claude agents — bookings, menu questions, complaints, all without human intervention.
Solo Engineer & Builder
Logistics dashboard with AI document extraction. Pilot built for a distribution company in Portugal — automating manual data entry across thousands of weekly delivery notes.
38 agents · multi-world orchestration
Solo Architect & Builder
A multi-agent orchestration system. 38 specialized agents running Claude Sonnet 4.6 in dedicated sessions, coordinated through a shared filesystem and an adaptive watcher daemon. Game-style HUD for human oversight. Local-first, no cloud dependencies, token economy enforced at the agent level.
Co-Founder & Engineer
The HR Operating System for Portuguese SMBs. Recruitment, documents, leave, compliance, and HR assistant — in one platform, designed for Portuguese SMBs from day one. AI woven through every module.
Production agents, structured outputs, long-context document parsing.
Used for: WhatsApp agent persona, tool-use over restaurant menus, JSON extraction.
Claude Sonnet 4.6 in tmux-based sessions, shared filesystem coordination, MCP integrations, token economy enforced per agent.
Used for: Fleet HQ — 38 agents in dedicated sessions, watcher daemon coordinating across worlds.
Vision models for document extraction, structured outputs, function calling — paired with Claude for redundancy and specialization.
Used for: DocFlow invoice extraction, vision-OCR pipelines, fallback model when latency matters.
Self-hosted workflows orchestrating multi-step automations at scale.
Used for: lead enrichment, multi-channel notifications, cron-triggered ETL.
Postgres, RLS, edge functions, real-time subscriptions.
Used for: customer state, vector search, row-level multi-tenant isolation.
Customer-facing AI agents at scale. Conversation state, webhook handling, message templates, multi-tenant deployments.
Used for: TheFacio — restaurants and bars running AI customer service end-to-end on WhatsApp.
App Router, RSCs, server actions, edge runtime — full-stack on a single mental model.
Used for: dashboards, marketing sites, edge-rendered SSR APIs.
Edge functions, KV stores, R2, Pages — global deployment without ops overhead.
Used for: webhook fan-out, file pipelines, geo-aware routing.
Strict mode end-to-end. Schema-first development. Zero 'any' tolerance.
Used for: every project — zod schemas at every boundary, never unknown shape.
Also working with
Production AI systems that ship. Claude, GPT, RAG, vision pipelines — built to handle real users.
Orchestration, coordination, agent infrastructure. Specialized agents working together against shared goals — with token economy and human oversight.
Operations that run themselves. n8n orchestration, custom backends, data flows across third-party APIs.
From idea to production. Next.js, Supabase, Cloudflare — full ownership, fast iteration cycles.
Contract part-time · 15–25h/week · Remote
I build AI products that ship. Not prototypes, not demos — production systems running LLMs in the loop: customer-facing agents, multi-agent orchestration, document pipelines with vision models. Based in Lisbon. Mostly self-taught, fully self-shipped.
I work embedded in small teams where decisions move fast and code ships weekly. I don't pitch — I deliver. If your roadmap is moving slower than your ambition, I'm the senior pair of hands on the AI / agents layer that fixes that.
