◦ Launched Snapchat’s first generative AI powered visual enhancement system for product advertisements◦ Drove development, data collection / labeling & training / fine-tuning process of Snap’s first in-house video captioner◦ Drove implementation of camera unlocking experience that enables AR content to be unlocked by scanning arbitrary images; used in thousands of lenses & high profile partner integrations as seen at the Super Bowl & Cannes◦ Launched visual classification via Scan Module in Snapchat’s Lens Studio which gives AR creators the ability to recognize 500+ object types across 4 categories (Objects, Places, Cars, Dogs) within a Snapchat lens◦ Drove development of Snap’s in-house emoji-aware OCR service used by scan visual search, platform integrity ads, spotlight and content moderation; drove wins across company, e.g. improved spam detection by 168%, reduced human moderation cost by $130k/yr. increased engagement bait detection by 160%◦ Drove integration of math problem detector and equation problem solver into Scan and MyAI product surfaces◦ Authored billion+ scale vector similarity search engine that powers visual search use cases (e.g. fashion, product, lenses) and content-based machine learning data collection pipelines (e.g. visually or audio similar media & cross-modal retrieval) used to collect candidate training data for 50+ ML use cases at Snap◦ Trained & deployed multiple vision models powering Scan including scene tagger, sky segmentation, hand detection◦ Contributed to backend & iOS/Android apps for context cards, widgets, search, messaging, friending & scan features◦ Managed backend and machine learning engineers; grew team from 0 to 10; promoted multiple engineers◦ Owner of machine learning infrastructure track / job family at Snap, defined job requirements and hiring criteria; owned cost goals for org and cut costs by millions of dollars per year; helped define LLM usage polices in engineering

◦ Developed software for product de-duplication / item authority system that continuously improves Amazon.com catalog quality◦ Contributed catalog validation logic & matching criteria; launched over 30 new Amazon.com product categories

◦ Conducted research in natural language processing & educational computing applications; resulted in 8 publications

◦ Designed experiments & developed text classifiers for use on military chat posts. Results published in ICMLA

◦ Conducted research in astrophysics & computer vision applications; resulted in 2 publications

◦ Developed web applications and tooling for supporting combinatorial software testing tool suites