Hi! My name is Duncan, and I'm an ML Engineer at Google. I'm interested in generative AI, causal inference, and optimization.
Connect with me:
About Me
I was born in Buenos Aires, Argentina and grew up in sunny South Florida. I love traveling, drinking mate, hiking, and anything Star Wars. I'm also a huge fan of the Miami Heat, the Florida Gators, and all Argentine sports teams!
Education
2022-2024
Master of Science (M.S.) in Statistics and Data Science at Stanford.
2017-2021
Bachelor of Science (B.S.) in Industrial and Systems Engineering at the University of Florida.
Work Experience
Current
I am a Machine Learning Engineer at Google, working on the Google Search team. I work on unstructured data retrieval using LLM's. This is part of Google's Proactive AI initiative, which aims to improve the search experience for users by providing more relevant and personalized search results.
Summer 2023
Working on the Google Search team, I implemented tree-based models for continuous retraining within Search Notifications. The model tunes notification cutoff thresholds daily to automatically meet daily target metrics.
This project is part of a larger effort to improve the Google Search Notifications experience for millions of mobile users, and is currently in production.
Summer 2022
Working on the Root Cause Analysis (RCA) team, I Implemented a change-point detection algorithm for isolating anomalies on time-series data from the Watchdog platform.
Additionally, I built a JS-based tool for visualizing time-series data and anomalies, and worked on improving the RCA platform's UI/UX.
Summer 2021
Working on the Google Looker Studio team,
I built a custom visualizations for the Looker platform, and worked on improving the Looker extension framework.
I also worked on a project to improve the Looker extension developer experience, and built a tool to automate the process of creating Looker extensions.
Management science research focused on understanding how employee sentiment towards their employer changes over time. Specifically, looking at Glassdoor reviews written by employees and comparing pre/post IPO sentiment across several topics.
This is a full stack project, and contains code for scraping the relevant reviews from Glassdoor, data processing, clustering and topic modeling, data mining and analysis.
Automating English Language Proficiency Assessments
Class project for CS224N: Natural Language Processing with Deep Learning. We built a model to automatically grade English language proficiency exams such as the TOEFL and FCE. We built transformer
models for written and spoken proficiency, aiming to provide a more accurate and fair assessment of English language proficiency.
Predicting Patient Fall Risk from Smartphone Videos
Class project for CS231N: Computer Vision. In an effort to design useful and practical models for simpler, cheaper,
and more accessible diagnostics compared to expensive laboratory tests, we built a model to predict patient fall risk from sit-to-stand videos filmed at home.
We used a combination of pose estimation and video classification to predict fall risk based on markers such as ankle angles and hip rotation.
