My career started in 2021 inside accounting companies — not software companies. Before I wrote much code, I was close enough to operations to see exactly where time was being wasted: analysts spending hours cross-referencing SPED files, accountants submitting the same forms manually for hundreds of companies, payroll teams maintaining GL mapping tables in Excel that depended entirely on individual knowledge and broke whenever someone was absent.
The first system I built reduced three hours of daily manual work per analyst. That result — people doing less repetitive work — shaped how I think about software permanently. Not the architecture. Not the framework. Whether the people using it have less to do by hand.
My work since has been a continuation of that same logic: find the repetitive process, understand why it exists, and build the system that makes it unnecessary. Payroll-to-accounting mappers. Government portal automation. NFSe pipelines. Certificate compliance tracking. SPED cross-validation. Firebird integration with legacy ERPs that have no external API.
The domain knowledge is not incidental. I understand what EFD-Reinf requires, how SPED files are structured, why the mapping between a payroll event and a GL account is non-trivial, what alvara submission means operationally. That depth changes what gets built, how fast, and how close it is to what the business actually needs.