HotMarcas: a trademark monitoring platform at national scale

HotMarcas is a Brazilian trademark monitoring platform that ingests, processes, and monitors intellectual property data at national scale. It automatically tracks trademark filings from the Brazilian patent office (INPI) and surfaces relevant activity to businesses. I built it end to end with Claude Code.

The challenge

Brazilian businesses have no easy way to monitor trademark filings that could threaten their brand. Official INPI data is published weekly in massive, hard-to-parse XML files. HotMarcas automates the entire pipeline: ingestion, parsing, storage, and monitoring at a scale of millions of records.

The approach

The core is an automated ingestion worker that parses the weekly INPI RPI XML publications using SAX streaming. Instead of loading a massive file into memory all at once, it streams through it, which keeps the memory footprint flat no matter how large the publication is. That is what makes ingesting millions of records every week practical.

The codebase is a monorepo that separates the web app from the ingestion worker, so the product and the data pipeline evolve independently. On top of the data sits a conversion-focused landing experience and a monitoring dashboard. The whole pipeline runs on scheduled jobs, pulling fresh INPI data every week with no manual step.

Tech stack

What this project shows

HotMarcas monitors over 68 million companies and 5 million trademarks, ingesting massive government datasets through a memory-efficient streaming pipeline that runs fully automatically every week. It was built end to end with Claude Code, from the ingestion worker to the dashboard: an AI-first product taken all the way to a live service operating at national scale.