A decade close to the machine, and not done

It started with taking computers apart to understand why one felt fast and another felt broken. That question never left. Razor began in 2014, in a garage in Passo Fundo, in the south of Brazil, funded by selling a car when I was 19. The first real sale was for my brother André, an architect who needed a machine that could actually run AutoCAD. Building the right one for him turned a repair shop into a company.

From there it became a craft. Match the system to the real workload. Validate it under the conditions it will actually face. Stand behind it with service that treats downtime as the enemy. A workstation maker from the Brazilian interior ended up competing with Dell, HP, and Lenovo, serving teams at Petrobras, Samsung, SAP, and Globo, and running one of the largest hardware equity crowdfunding rounds in the country.

Selling to people who could not afford to be let down raised the bar in a clarifying way. Performance was never a spec sheet. It was context, workload, thermals, drivers, bottlenecks, and trust. After enough machines, one pattern became impossible to ignore. The hardware was rarely the real problem. The missing piece was the machine understanding itself.

I could build better hardware forever and never close that gap. So I started building the thing that could. Today I work at the intersection of systems performance, local AI, telemetry, and human-centered computing. The goal is simple to say and hard to do: turn the signals a machine already produces into something a person can act on, privately, in the moment. I am currently a Founder in Residence at Antler in Singapore, building the next chapter from there.