Data Science Professor. I'm teaching a semestral data science course which covers the following topics: Python fundamentals for Data Science (Numpy, Pandas and visualization libraries), SQL, Descriptive and inferential statistics, data cleaning and preprocessing, APIs and web scraping, Machine Learning fundamentals, Supervised and unsupervised machine learning algorithms, Recommender Systems and Text mining.
Etermax is a mobile game development company, mostly known for the game Trivia Crack. My main projects at this position are the following: Implemented an image moderation system with CNNs using fast.ai. Explored various improvements on the question serving process using state-of-art and old fashioned NLP, clustering techniques, Pyspark and various ml models (question recommendation strategies based on collaborative filtering, trivia embeddings for detecting duplicates, similarity models for increasing content diversity). Implemented a simple rule-based categorization system for questions with high coverage for more than 100 classes. Currently, we are working on a win-rate prediction project for content personalization, trying factorization machines (FFM), collaborative filtering (Pyspark ALS, LightFM) and classification approaches. Technology stack: Pytorch, fast.ai, transformers, Spark, Databricks, AWS (EC2, S3, SageMaker, Lambda), Python Data stack, Redshift, MySQL, mlflow, airflow.
TechnologiesPythonSQLPyTorchPySparkRecommendation SystemsAWS EC2
Navent is the leading online classified ads company for jobs and Real Estate in Latam, with a presence in 8 countries. Some of my main achievements at Navent are the following: Implemented a prod2vec item-item recommendation model from scratch and scaled it up with k-means, annoy and multiprocessing obtaining a 17.5% increase of CTR against previous model. Scaled-up the principal recommendation system to handle 10x its original data using spark, annoy, Cassandra and GCP obtaining a 30% increase of CTR against previous model. Did reporting with Jupyter notebooks and Google Data Studio and segmented users using clustering techniques. Participated in a duplication detection system project using phashes, Mongo, Cassandra and GCP. Improved and scaled the recommender events tracking system using PubSub, Dataflow and BigQuery. Presented our recommendations ecosystem at Universidad Austral and at UADE (austral.edu.ar/ingenieria-posgrados/jornadadm/Navent-Sistemasderecomendacion.pdf).
TechnologiesPythonSQLGoogle CloudSparkRecommendation SystemsBigQuery
Worked abroad (in Israel) for almost 2 years, remotely from B.A. after that. I was part of various teams involving members from different countries (Israel, US, India, Ukraine and China) handling different time zones and levels of expertise. Worked mostly on building a CI system from scratch with my main team, but we also coded a few utilities for the main C developers.
Back-end developer. Worked mostly building, deploying and maintaining an e-mail delivery system from scratch with PHP. I learned a lot about development, databases, Linux and requirement analysis in this position.
Simple CRUD web-based apps with PHP.