Case study

MercadoLibre

Largest online retailer in Latin America.

About

MercadoLibre hosts online ecommerce and auction platforms that provide users with buying and selling mechanisms for etransactions.

MercadoLibre
Enterprise Software

FOUNDERS:	Marcos Galperín, Hernán Kazah
STOCK:	NASDAQ: MELI
LOCATION:	Buenos Aires, AR
TIMELINE:	2015 - 2017
SERVICES:	Machine Learning & Management UI.
OUR TEAM:	7
WEBSITE:	http://www.mercadolibre.com

Our Mission

MercadoLibre, a company with more than 100M users and over 900 IT employees in 12 countries, established a working relationship with us in order to solve issues related to the category tree of their immense product catalog.

An iteration based on specific goal achievement, ongoing development and progressive team scaling are the basis from which the solution is being crafted.

The problem itself brings three main tasks: Research & Development, bringing into production and exposing it through an API and Frontend development..

Main challenges

State of the art

As it is a very specific problem that we are dealing with, there is lack of readily available solutions to build upon. Constant research and a lot trial & error are needed to build the most accurate solution.

Data

Immense amounts of data need to be archived and correctly analyzed.

Development

R&D implied a massive historical data analysis which later directed us to the hypothesis, trial & error circle until a fair solution was found. At last, we ended up building an algorithm which uses neural networks and users' navigation data in order to facilitate the handmade categorization. We also built a set of algorithms using deep learning and NLP tools in order to automatically suggest missing categories. Some of the tools used where Gensim, Keras and NLTK.

Added to this, as human validation is part of the solution, we built an ultra intuitive and intelligent frontend using React + Redux, ES6 and Webpack to communicate with the built API.

Through ongoing development, we achieved results on a sprint basis. The most relevant result is that categorization has become much more efficient and we implemented some automation of category creation and product classification.

Read more about Mercadolibre on Wall Street Journal, Yahoo Finance & New York Times.