Founder by experience. Technical builder by curiosity.
I have spent years building companies, investing in others and working inside Brazilian fintech and credit regulation. Somewhere along the way I got curious enough about how the machine works to start building it myself: AI agents, the systems that keep them honest, and the small tools I need along the way.
I am not a career engineer, and I am not starting from zero. I am a founder who builds.
Based in São Paulo, Brazil.
- How do you know, for real, that an AI agent is doing good work?
- How do you govern a team of agents, and the people around them, without slowing everything down?
- How do complex systems like credit, regulation and operations turn into simple products?
- What does endurance teach about building companies? So far: steady pace beats the isolated sprint.
- GovEvo: the system where my agents operate, evolve and are governed day to day.
- KODA: an AI agent for e-commerce over WhatsApp, under the House of Pace (HoP) brand.
Both are private.
This profile is the person and the questions. The repositories are the experiments and the evidence.
- llm-council: several LLMs from different providers answer the same question, review each other blind, and a chair outside the council synthesizes. CLI and MCP server.
- papers-journal: a daily journal of the Hugging Face Daily Papers, triaged against a declared interest profile.
- DataJud: a small local client for the public DataJud/CNJ court-data API that keeps the raw response and the provenance of every query.
- macro-br-juros: Brazilian interest-rate and implied-inflation curves as pure TypeScript functions.
- building-in-public: where notes on what I build, what broke and what I learned will go.
I take cycling as seriously as dashboards: if it can be measured, I am measuring it. Long rides and long-running companies ask for the same thing, a pace you can hold.