PhD Candidate @ UMich · gamba-remove-@umich.edu
Hi I’m David! I’m a PhD student in Information Science at the University of Michigan. I work with Daniel Romero and Grant Schoenebeck. Previously, I worked for a few years in the field of Analytics and Data Science. I worked in positions from consulting to data engineering to model developing, in applications mainly in forecasting and recommendation systems. I came back to academia to develop research on my interest in data science for social good and for understanding societies and groups.
My research centers on ML and AI for collective consensus, preference modeling, and decision-making. I'm also looking to apply these methodologies to social contexts such as content moderation, text consensus and participatory democracies.
I’m fairly active in participating in projects that use data for social good and I’m always interested in discussing and developing ideas on that space. Other areas I like to experiment a lot are cooking and coffee making. Also happy to chat about this.
Nice to meet you!
Also on Google Scholar.
Reading and Writing group to study algorithms and AI from a Political Economy and critical perspective. Officially a Rackham Interdisciplinary Reading Group.
Volunteered in Data Science project with the NGO Omdena to build recommendation systems based on NLP models to identify common psychological violence information resquest and reply with the appropiate resources.
Volunteered in Data Science project with the NGO Omdena to use satellite imagery to extract possible malaria sites to aid the mission of the startup ZzappMalaria.
Lead and mentor technical research on methods to improve granular forecasting across millions of items in Mercado Libre's inventory. Led the research and implementation of metrics for the development of forecasting methods aligned with business objectives
Developed data infrastructure and pipeline architectures to process text analytics and computer vision applications in the company. Optimized machine learning models to run at scale and integrated these models in the data pipelines
Develop machine learning ranking models used in the core of the business improving 10% over previous performance metrics. Facilitate data-related outreach activities to promote data science and Merlin, including giving talks at seminars, lecturing workshops of applied machine learning and organizing outreach events.
Developed models for sizing of the sales agent workforce in order to reach monthly sales and margin goals. Developed a forecasting system for sales KPIs with models under 20% error, together with reports and notifications. Analyzed the patterns of returning customers and implemented a model to predict the time of return for appropriate timing of calling frequencies to clients from the contact center.
Engaged with customers to understand their business objectives, analyze their data and translate said objectives to Data Science requirements and projects. Also developed these data products, for EDA visualization and dashboards with Dplyr, Pandas, ggplot, Altair and PowerBI and for machine learning models using R, Caret, mlr, Python, Scikit-learn, Pytorch, keras and SparkMLib. The cases included churn prediction, chatbots, inventory management, demand forecasting and client characterization.
Under supervision of Ana M. Rey, we studied the theory of entanglement measures such as Spin Squeezing, Quantum Fisher Information and Entanglement entropy in the all to all Ising Model, with calculations and simulations performed in Wolfram Mathematica. We conducted this research with the objective of finding new benchmarks for the current experiments with ultracold atoms at NIST, which are used in the building process of quantum computers.
Developed new course material for courses in computer infrastructure, in particular labs, for topics in low level software concurrency, performance and security. In addition, I lectured the lab sections and assisted students with all doubts regarding the course, and guided their project developments.