About

I have a master's from ITA (Instituto Tecnológico de Aeronáutica) and spent the past decade building ML and data systems in production. I was part of the founding team at Pipo Saúde, worked at Loggi and Elo7, then moved to Meta and Expedia where I built recommendation systems and AI pipelines serving millions of users.

Along the way I kept running into the same thing: most AI projects never make it to production. The model works, the offline numbers look great, you ship it, and nothing happens. I got really interested in why that keeps happening. It's usually not the model. It's the data, the systems around it, and the way teams are set up.

That's what I do now at Dipolo AI. I work with companies to find and fix the data problems that are quietly killing their AI projects.

I write to think. Most posts start as notes I take while debugging something, building a company, or just trying to make sense of an idea. Writing forces you to be precise. You can't get away with a vague understanding when you have to explain it from scratch.

Topics range from applied ML and AI systems to startups, entrepreneurship, and the occasional story about Gaia, my dog, who was once featured by the BBC as a [genius dog](https://www.bbc.com/portuguese/articles/cpq9lw4yqxxo). I write about whatever I find interesting or worth thinking through.

I write in both English and Portuguese. Some posts appear in both languages, others only in one, depending on who I think will get the most out of it.