Success story: Mercado Libre

Discover how we successfully optimized fraud and irregularity detection on the platform and trained its Machine Learning model.

The challenge

Mercado Libre is an electronic commerce and payment platform that runs in more than 18 Latin American countries. On this platform, almost 73 million people interact, between buyers and sellers. In 2020 alone, it generated the delivery shipment of 1,7 million packages per day and added 57 million new users. The company has managed to position itself as a leader in the region by guaranteeing a reliable ecosystem for users. However, the growing volume of data and exchanges in the marketplace made them turn to Arbusta: Mercado Libre needed an external team that could improve image and text labeling to help its Machine Learning model prevent frauds and verify the quality of transactions.

The solution

Our Machine Learning Training team took on the manual training of the model supervised by Mercado Libre to help it become more accurate in the detection of frauds and irregularities. We initiated a process of manual labeling of the platform's images and texts to assist the algorithm in verifying the authenticity of logos (image annotation), identifying publications that violate counterfeit policies or involve forbidden articles (fakes), verifying the accuracy of information provided by sellers to buyers, and ensuring correct filling of text fields in each publication.

Others successful stories