Moves

Mobile Application

NLP

Social Media Influencer App

A virtual assistant mobile app to assist celebrities in increasing their social media reach/influence on popular platforms.
Provides real-time assistance while posting content by giving suggestions based on content sentiment, likely impact and upcoming events.
Generates influencer score based on user popularity.
Provides opportunities for brand association based on the range of influencer scores.
Verify’s and validate’s brand-related goals performed by the user.
Web dashboard/graphs to track the growth of user popularity and other stats.

Bulk storage of raw data done in Amazon S3.S3 enables fast retrieval and provides advanced querying capabilities over bulk files.
Analytics over raw data is performed on the Amazon EMR cluster using Apache Spark. Spark on EMR enables us to distribute the processing load across a cluster which can be scaled up/down based on the requirements.
User content sentiment is predicted to be positive/negative/neutral with the help of models generated using TensorFlow.
Brand-related goal/logo verification is performed by using image detection networks(FRCNN) based on TensorFlow.
A Linear Regression Model is used to calculate the influencer score of each user based on various factors.
MongoDB acts as a fast-serving layer to store the final results and provide fast querying capabilities to the application/dashboard.
The Android notifications were handled using FireBase FCM, the iOS notifications were handled using APNs, and the web notifications were handled using Sockjs.