> I'm a math and music nerd currently at Spotify as a Research Scientist, working on large-scale recommendation systems and generative sequential modeling. I recently earned an MA in Statistics from The Wharton School and an MEng in Systems Engineering from SEAS at the University of Pennsylvania.
Before graduate school, I worked on genomic motif discovery through ATAC-seq analysis at Stanford's KundajeLab. I also interned at CERN
, contributing to the CMS Open Data initiative alongside Dr. Lassila-Perini.
Originally from the beautiful city of Montevideo, Uruguay , I studied at Universidad de la República. During this time, I played for Uruguay's national Waterpolo
team and worked as a software developer at IBM. If any of this sounds interesting, or if you want to share some mate
, feel free to reach out.
Broadly, I work on developing reliable learning systems that scale effectively with data and compute. At Spotify, I apply these methods to large-scale recommendation systems, with a particular focus on sequential generative modeling.
Throughout my graduate studies, I adopted a data-centric perspective of ML, focusing on continual model fine-tuning as new, diverse data emerges. This has driven my work in active and continual learning using large pre-trained models. On the theoretical side, my work on duality-based constrained optimization has shown that dual subgradient methods can yield near-optimal and near-feasible solutions, without randomization, despite non-convexity. This allows models to not only excel at their main task, but also meet critical requirements such as robustness and fairness.
I also have a particular interest in problems involving biological signals such as genome sequences, medical images and the gut microbiome.
Near-Optimal Solutions of Constrained Learning Problems
Juan Elenter, Luiz Chamon, Alejandro Ribeiro
International Conference on Learning Representations (ICLR), 2024
LoRanPAC: Low-rank Random Features and Pre-trained Models for Continual Learning
Liangzu Peng, Juan Elenter, Joshua Agterberg, Alejandro Ribeiro, René Vidal
International Conference on Learning Representations (ICLR), 2025
Feasible Learning
J. Ramirez * , I. Hounie * , J. Elenter *, J. Gallego *, A. Ribeiro, S. L. Julien
AISTATS, 2025
A Lagrangian Duality Approach to Active Learning
Juan Elenter, Navid NaderiAlizadeh, Alejandro Ribeiro
Neural Information Processing Systems (NeurIPS), 2022.
Some stuff I like about Uruguay.
In the words of one my engineering professors: "To make my heritage more informative, let me tell you that Uruguayans are idiosyncratic folk. Having a working knowledge of Uruguayans means you are aware that we are obsessed with fútbol (soccer) and have managed to stay within the global top-10 despite the low population number. You are also aware that we drink mate from a calabash (a type of gourd) shell instead of tea or coffee from a mug and eat 125 pounds of beef per year. That comes at about a 6oz portion per day. Except we think a portion of beef is a full pound which means we eat beef only once every three days. On a more serious note Uruguay is very socially liberal. It was one of the first countries in the world to enact laws for free public education, 45-hr workweeks and female voting at the turn of the twentieth center and among the first to legalize marihuana and same sex marriage at the turn of the 21st. A certain anarchist vein runs strong. We are suspicious of government and have an obstinately uncooperative attitude towards authority. We nonetheless have a very large government sector because as much as we are suspicious of government we are more suspicious of corporations. Disregard for authority extends to divine powers. Half the population defines themselves as agnostic or atheist and active practice of religion is rare. What Uruguayans call “the right” is to the left of Bernie Sanders."
Water is my natural environment.