Designed, built and maintained several batch and real-time data pipelines:- Based on Spark and AWS EMR clusters to process TBs of data and derive analytics.- Based on Apache Flink to process GB/s of data and derive information to be fed back into our products.- To do ingestion from several sources, including web crawlers and APIs, in order to feed our products with additional intelligence.- To estimate the approximate GPS coordinates of events identified by our products. This was done by triangulating some of the data in the events with information we maintained in a graph-based database.- To feed data lakes based on AWS services (S3 and Athena) and provide analytics dashboard (based on quick sight).

Worked at a data science consultancy company. My main responsibility was to work alongside the data science team in extracting insights from a dataset belonging to one of the largest perfume companies in the world. We were given a dataset of perfumes from which we had to generate meaningful clusters. While I collaborated in the research process I focused mainly on providing supporting infrastructure, for example, coding algorithms and data ingestion/processing pipelines in Spark. Afterwards I was charged with creating a front-end based tool that allowed us to visualize the graph structure of the clusters alongside other meaningful statistics.

Worked in several machine learning projects, mostly focused on recommender systems. These required data analysis, model training/selection and deployment into a production environment. In chronological order:- A web app for literary quotes. To complete the project I had to scrap quotes from a popular social network for book readers, clean and format the data and decide which recommender model to use. I had to put everything together by creating a website that allowed users to register and receive daily quote recommendations by email, which they could rate so as to improve their personal recommendations.- Hotel recommendations and automatic static ad creation. I was tasked with creating a system that generated static ads for hotel rooms on demand by passing a zipcode. The app would query an API for hotel rooms around a zipcode and organized the options according to several criteria. Finally, it would render images of the ads by taking a screenshot from a headless browser. I had to work alongside a web designer to generate the themes for the ads. The system was successfully used as a prototype for a business pitch.- Product recommendations for an online only grocery e-commerce. I was given a dataset of products with its metadata. I selected the features to use for the recommendations and created a simple content-based recommender which I put in production and made available through an API.- Created a prototype artwork recommender system which used the procedures outlined in this paper (https://arxiv.org/pdf/1505.00855.pdf) to extract features from images. The paper uses a combination of three sets of features which are then reduced by PCA and finally used to train a similarity metric. The recommendation was based on finding similar artwork, according to the metric, to the ones the user had liked.

Worked creating visualizations and analytics that helped us gauge the performance of our ads. Generated several tools to automate processes for various departments. Performed client integrations, a key step in the process of automatic ad creation.

Maintenance and updating of tools that helped improve the IT team’s productivity. I suggested improvements in the infrastructure, they were discussed, accepted and began their implementation.

Responsible for the finances of the organization, as well as in charge of raising the funds necessary to support the SABF team operations.

Creation and maintenance of tools that supported the operations of the organization (www.sabf.org.ar). I helped migrate all our codebase from PHPCake to Python Django and coded new tools in order to automate processes that allowed the organization to scale.