Machine learning for E-commerce


Falabella e-commerce is the largest e-commerce in Latin America. The amount of data available makes any data scientist face new challenges and opportunities for learning and helping the business.

From 2016 to 2018 (sporadic work) I developed and implemented machine learning systems for Falabella. Working as a Data Scientist and a Web Intelligence Engineer, I learned how to deploy machine learning models and their outputs. My team also faced many challenges on data-architectures that we helped to solve.

I led a data scientists team to outperform the user experience and findability at from a machine learning perspective. At the same time, I was part of the team that evaluated and defined the new Big Data technologies and protocols at the company. I also spread a data-driven mindset through the company and shared my scientific knowledge with my colleagues.

It was a great honor and opportunity working with them parallel to academia.

My main contributions were:

  • First data scientist at
  • Lead a data scientists team to outperform the user experience and findability at from a machine learning perspective. In charge of defining the project schedule,  needed resources (technology and people-networks). The project covers a full CRISP-DM scope, including business and data understanding, data integration from several sources and systems (on-premise and cloud), training and testing machine learning models (recommendation, clustering, ranking, and classification), and deployment through different strategies. 
  • Designed and implemented the theoretical and technical tests used to hire data scientists at and other companies of the holding.
  • Recruited and interviewed data scientists for the company. It included various profiles: data wranglers, data engineers, data scientists with mathematical and algorithmic mindsets, and data science team leaders.
  • Participated in the evaluation and definition of the Big Data platforms and protocols for the Falabella holding.
  • Started to set up a data-driven strategy and mindset through the company, and motivated by sharing my scientific knowledge with my colleagues.

Cencosud commerce

During 2016 and 2017 I developed two consulting projects for the Customer Intelligence department of Cencosud.

  1. Clustering of customers according to their habits
  2. Recommendation of discounts for customers

The main challenge was to understand the buyers' behavior and the business insights.

Perfect Match - Positive and negative feedback

Footwear recommender system

I developed the recommender system behind this app (only available for mobile devices).

Falabella had about 4000 sandals to sell, but its customers had no enough time to scroll all their pages. I developed a recommenders system that tries to converge as fast as possible to the best product for each customer.

This app was launched as an MVP during 2018.

Related keywords - Recommended queries

Using navigation data, I implemented a model that recommends words to specify search queries. This way the clients learn how to use the search engine better and find their products faster.

This works parallel to the type-ahead functionality, which I also helped to set up.

This tool was launched during 2018.

Products on development stage

I left Falabella before that all my team's projects were fully deployed. Here I specify some of these projects:

  • Relevancy score: Researched on the preferable model to increase relevancy. Designed the full pipeline to deploy that model, including all the stages of a Data Science project.   This project is currently deployed in its first stage and showed to outperform all previous techniques of the relevancy score at

  • Products' bundles: Implemented the first version to offer the  "customers who bought this item also bought that..." tool. Trained in transactional data.