Built an online image-scraping and computer-vision processing pipeline

Research Engineer · New York, NY · 2015 – 2017

Scrapy spiders crawled online sources for images of bodies, filtered them and saved them with their metadata to S3, run per source (an Instagram spider by hashtag and count, for one). The crawl then ran through the images pipeline, which filtered the body images on a GPU EC2 instance ahead of training.

Asset Pipelines · Computer Vision

Python · MongoDB · Docker

← Animated FBX from body-capture data, with a real-time Unity client
→ Built BodyLabs Blue, an apparel size-recommendation app for e-commerce checkout