About

Most marketers think in stories. Most engineers think in systems. I've spent my career in the gap between those two — where you need intuition about what moves people and the rigor to prove whether it actually did.

I grew up during the startup wave in São Paulo. My first jobs were at places where everyone did everything — I'd be writing SQL queries in the morning and designing ad campaigns in the afternoon. That forced generalism turned out to be the most useful thing I ever learned, because it trained me to see the same problem from both sides: the creative side that asks "will this resonate?" and the analytical side that asks "how would we know?"

I spent a few years doing digital marketing for Nike in Brazil, then seven years at Meta across Facebook and WhatsApp. At Meta I worked in São Paulo, then moved to the Bay Area. The thread through all of it was the same: I kept ending up where storytelling met measurement — shaping how a product was perceived in one conversation, then building the system to prove whether the perception changed anything in the next.

Working across twenty-plus markets — US, Brazil, India, Germany, the UK — gave me a permanent outsider lens. You learn quickly that what works in one culture fails in another, and that the interesting insights come from the gaps between how different markets see the same product. I still think about marketing problems that way.

After Meta I did stints at Qualcomm, Intuit, and HCLTech. Each one taught me something different — semiconductors, data infrastructure, enterprise AI. Now I'm at ResMed doing digital marketing for their Growth & Incubation team in digital health.

How I got to AI agents

The same instinct that pulled me toward measurement eventually pulled me toward building. At Meta I kept needing tools that didn't exist — ways to see what was working, connect teams, automate decisions. So I built them. I never planned the transition from strategist to builder. It just kept happening because the problems I cared about couldn't be solved with a slide deck.

AI agents clicked when I noticed the hard part wasn't the model itself — it was the design problem around it. How do you break ambiguous work into pieces that can run in parallel without falling apart when you put them back together? That's orchestration. I spent a decade learning it by managing campaigns across markets and cross-functional teams before I realized the same thinking applies directly to AI.

What I think about

I'm interested in what happens when execution becomes nearly free. Models are capable enough. The bottleneck is the human judgment around them — what to build, how to scope it, when to trust the output.

I also think a lot about what cheap execution means for expertise. If AI handles most of the building, does the person who spent years developing skill become more valuable or less? I don't have the answer. I write about the question because I think it matters more than the industry's current optimism suggests.