serenata de amor

The project analyzes reimbursement claims filed under the CEAP (Quota for the Exercise of Parliamentary Activity)covering expenses like meals, travel, and office suppliesand flags suspicious transactions that appear to violate reimbursement rules.

It is one of the earliest and most prominent examples of civic AI applied to anti-corruption transparency work.

The project's core tool, Rosie, is a machine learning model that processes thousands of reimbursement records monthly, identifying patterns consistent with irregularities such as duplicate claims, expenses at closed businesses, or amounts far exceeding reasonable norms.

When Rosie flags a suspicious transaction, the project publishes the finding publicly on social media, creating accountability through transparency and enabling journalists, civil society organizations, and citizens to follow up with formal investigations or parliamentary inquiries.

Serenata de Amor has been cited as an influential model for applying open-source AI to government accountability challenges in emerging democracies with limited institutional oversight capacity.

Its methodologycombining public government data, open-source Python ML tooling, and public communicationhas inspired similar projects in other countries.

The project is developed collaboratively by volunteers from the Brazilian tech community and has contributed meaningfully to both recoveries of public funds and broader awareness of how technology can strengthen democratic institutions.