I build systems that don't fall apart in production.

Both. As an AI engineering consultant and fractional technical lead, I embed with teams that need senior architecture capacity without a full-time hire, owning technical decisions, unblocking developers, translating tradeoffs for stakeholders. I also work directly with founders and operators on specific scoped projects: RAG systems, automation pipelines, WordPress AI integrations.

A free 30-minute Intro Call. It answers one question: is your project something I should take on? If it is, the next step is a paid diagnostic: I look at the code, the logs and the systems involved, and you get a written summary with a recommended scope, a price and the assumptions both depend on. You can book the Intro Call directly, or send your scope through the contact form first.

Access to whatever the work touches, from day one: the codebase or repository, a staging environment, server or hosting access, and API credentials for the systems involved, like the payment gateway, the CRM or the AI providers. I can't diagnose what I can't see. I also need a decision-maker who answers when something blocks progress.

A working system and the documentation to run it without me: architecture decision records explaining why key choices were made, integration guides, operational runbooks for common failures, and onboarding notes for the next developer. If you want ongoing support after handoff, it's scoped separately, during São Paulo business hours, with a reply within one business day and within 4 hours for urgent issues.

An AI systems architect designs and builds the infrastructure that makes AI usable in production: RAG pipelines that retrieve accurate information at scale, LLM integrations connected to real business systems (CRMs, CMSs, payment gateways), and automation workflows that replace manual processes reliably. The role bridges software architecture and AI engineering, not building demos, but systems that handle real traffic, real edge cases, and real data.

Most AI projects fail in production because they were built for a demo environment: clean data, no concurrency, no error states, no cost constraints. Production systems deal with malformed inputs, API rate limits, ambiguous queries, and unexpected load. Bridging that gap requires treating AI components as infrastructure, with the same logging, fallback handling, and architectural discipline as any production service.

A RAG (Retrieval-Augmented Generation) pipeline retrieves relevant documents from a knowledge base at query time, then generates an answer grounded in that retrieved content. You need one when accuracy matters: regulated industries, support systems with proprietary product knowledge, or any context where a hallucinated answer has real consequences. A stuffed system prompt breaks down at scale; RAG keeps the knowledge base separate, auditable, and updatable.

Yes. Typical integrations include AI-powered customer support chatbots connected to your product catalog and order history, content generation pipelines that draft posts for human review, and n8n automation workflows that replace manual weekly processes. The work connects to your existing infrastructure, no platform migration required.