I learned a great deal about previously unfamiliar topics while working on Emerge, from computer vision to systems architecture and more. Emerge was first conceptualized as an experimental venture in utilizing Computer Vision and NLP to organize disorganized social media data into highly specific search queries — for example, "accounts that are pregnant or have given birth in the past 2 years" ...
With over 5 million accounts in its database, processing over 6 million posts per day, the sheer volume of data that has to be fetched, processed, stored, updated, and retrieved on a regular basis presented big challenges across both development and service delivery/experience — how might we build a data-intensive AI application that is able to deliver a consistently accurate, reliable, scalable, and manageable experience?
Working closely with engineers, data scientists, DevOps, and customer success, product requirements were ruthlessly prioritized and developed in agile methods, for maximal impact per unit of resource investment and 'Time Value of Shipping'. For example, account search queries with the biggest market demand, applicability, and differentiation were prioritized for highest accuracy and data volume.
—
By leveraging NLP and visual AI, Emerge is able to make sense of over 5 million Instagram accounts in its database — down to what’s being discussed or what visual content is contained in individual posts. This level of in-depth understanding powers Emerge’s KOC Search experience, allowing users to search across 12 criteria and 32 pre-built personas.
