We like to tackle the hardest challenges first while keeping the entire scope of the project in mind. That challenge would be the two different artificial intelligence models we create to tag keyword metadata to photographs. Once a proof of concept demonstrated an average 60% accuracy, we moved on towards the business intranet side of the company (the models can be retrained and fine-tuned in parallel). There was a massive undertaking after thousands of photos were taken during events to tag each one manually then distribute them to their Dropbox folders. We developed a distribution system to automatically populate the customers' folders with their images. Simply upload a .zip file of everything and the system populates customer folders instantly.