Engineer first. Consultant second.

The short version

I've been writing code professionally since 1998 and haven't stopped. The problems got bigger, the tools got better, the fundamentals didn't change much.

I spent most of that time as an engineer, not a manager — and most of it working with data at scale, from fintech payment rails to petabyte-scale ad pipelines. For the past two years I've been integrating AI agents directly into production data engineering workflows.

That's the whole pitch: data engineering and AI integration, together. Because AI without solid data engineering is just an expensive demo, and data engineering without AI-aware tooling is about to feel very dated.

The path here

1998 — Brazilian media conglomerate (5+ years)

Large-scale web engineering when that meant Tomcat deployments and hand-rolled caching. I learned early that engineering teams at scale are mostly organizational problems dressed up as technical ones.

Telecom software, Germany

First international role. The specifics changed — protocols, process, office politics — but the engineering mindset didn't. That portability has shaped every engagement since.

Brazilian fintech — payment processing

Nothing sharpens your engineering instincts like systems where a bug moves real money the wrong way. I've been careful about side effects ever since.

Brazilian e-commerce marketplace

Kubernetes-based CD platform for the company. I also led a bug-squash hackathon that cleared 70% of the backlog in a week. The lesson wasn't about hackathons. It was that permission — not tooling — is usually the bottleneck.

Major web services provider

Systems migration and GDPR compliance. The kind of work no one writes blog posts about, but the kind that keeps the internet running. I got quieter and more careful here.

Top-10 global ad tech platform (4 years)

Data team, pipelines at 1.5 billion records per day, petabyte scale. This is where I stopped thinking about engineering as making things work and started thinking about it as making things work cheaply, reliably, and at volume. Most of the economics intuition I use today came from these four years.

Leading global travel platform

Built the company's new data pipeline and data lake on AWS — S3, EMR, Glue, Redshift, Lambda. The main outcome wasn't the platform itself. It was unblocking the ML team with near real-time data without impacting production. Giving other people leverage is underrated. It's usually the whole point.

Fast-growing logistics startup

Data platform from zero. Lake, warehouse, orchestration, analytics, team. Some decisions I'd make the same way again; some I'd redo. That's the zero-to-one tax — paid once so you don't pay it again.

Major B2B data platform (current)

Pioneered integrating AI agents directly into our data engineering workflows. Not demos. Not agent-framework experiments. Production pipelines where LLM agents propose changes, write code, and get reviewed like any other contributor. That work — and what I've learned making it survive contact with production data — is why I'm available now as a fractional architect.

Why data & AI, together

Most consulting offerings pick one. I don't, because of a specific observation: in practice, the teams getting real leverage from AI aren't the ones with the best prompts. They're the ones whose data foundation — pipelines, schemas, knowledge bases — was designed with AI access in mind.

The teams struggling are the ones with ambitious AI goals bolted onto weak data engineering. You can't have one without the other. So I help teams build both together.

What I actually do

How I prefer to work

Three formats, all remote-first and designed for international collaboration:

• Advisory — a senior second opinion on architecture, AI strategy, or tooling decisions. 2–5 hours a week.
• Fractional — I join your team part-time and ship production work alongside your engineers. 10–20 hours a week.
• Build — scoped, fixed-deliverable platform or AI-integration engagement.

More detail on the services section of the homepage. I work with small-to-medium engineering teams. I don't pitch. I ask questions, listen, and tell you honestly what I'd do. If the engagement isn't serving your team, I'd rather you know in week two than month six.