Khas Apke Liye
Description
A content-based recommendation engine to serve better & personalized news articles on the website of Dainik Jagran Hindi daily having 31M+ readers per month
Responsibilities
- Extracting data from newly published articles using CMS API
- Training a content model on the corpus of Hindi articles & building user preferences
- Deploying the model on GCP for serving the recommendations as JSON
Achievements
- Achieved >8.9% CTR with the highest being >15.0% on mobile devices
- Improved AvgSessionDuration by >5 minutes
Technologies Used
- BigQuery to create user profiles using user’s session data from Google Analytics and GA Realtime Reporting API to extract realtime user session data on client’s websites
- Jupyter Notebook, DataFlow to explore and process data and Gensim to train topic models
- Google Kubernetes Engine and Google Cloud Function to serve recommendations
- Google DataStudio for performance reporting
Libraries and Frameworks Used
- Numpy, Pandas, Scipy, Gensim, FastAPI