Conde Nast

Oct 5th, 2018

Conde Nast is one of the largest private media companies in the world, owning brands such as Vogue, GQ, Wired and Vanity Fair. They had a large data problem on their hands.

I briefly worked with them to assist in the delivery of a automated Generative AI fashion show tagging enrichment pipeline.

Background & Objectives


Conde had millions of imagery, from fashion shows, runways, and other sources. It was hard for employees to find the right images they need without spending many hours sifting through thousands of them.

The challenge is to streamline this process and make it significantly easier for editors to find the correct images they needed for their articles.


To successfully deliver on the requirements set, at a high level, the objectives outlined were:

  • Create a GenAI system to organise and search millions of fashion show, runway, and other imagery.
  • Reduce the time employees spend searching for images by 70%.


GenAI Tagging Enritchment Pipeline

In order to consolidate all images into a central repository, we began with data engineering and resolved immediate issues by streaming images to different cloud AI Image Recognition providers to tag and process images.

Post processed image data was then stored alongside the image inside a graph database, ArangoDB. Multiple pipelines were created for image ingestion and re-tagging, and we implemented AWS Elasticsearch on top of the DB to enable users to quickly search through the tagged images.

Lastly the frontend was integrated into their existing systems via an API to enable authors to use it.

Tools & Technologies

This project branched into many varying aspects of engineering and technologies.

  • HTML5
  • CSS3 (SASS / LESS)
  • JS (Vanilla / OO)
  • React
  • NodeJS
  • Architecture
  • PHP
  • MySQL
  • WordPress
  • GitHub
  • SSH
  • SVN / GIT
  • Bash
  • Python
  • AI
  • Data
  • AWS
  • Azure
  • GCP
  • Tensorflow
  • Google AI
  • SublimeText 3
  • Microsoft Office Suite