Raviteja Uppalapati

Senior Manager, Data Science

Data Scientist turned Manager at American Express with 7+ years of experience in building scalable text based algorithms and digital products. Interested in the application of machine learning and computational techniques for material and drug discovery.

Quick Links

CV
Interests
Projects
Courses

Brief Timeline

Skills

  • Python
  • Hive/SQL
  • Hadoop
  • Spark
  • Tensor Flow
  • Scala
  • Matlab
  • HTML

Projects

American Express, New York
Recommender Systems image error
Developed a recommender system using Restricted Boltzmann Machines (RBMs) to improve relevance of Amex merchant offers that resulted in a lift in precision of 15% over an existing technique.
Data Integration image error
Using Bayesian inference, built a merchant database by integrating conflicting information from multiple structured data sources like Yelp, TripAdvisor, Foursquare, D&B etc.
Hierarchical Classification image error
Built a hierarchical classification model using an ensemble of gradient boosting and pseudo relevance feedback to classify merchants into a 30,000 industry categories. Created features using word embeddings on text data scraped from web search results.
Location Intelligence image error
Developed a multi-language address parsing tool using neural networks that resulted in an improvement of 12% in geocoding accuracy. Created geo-visualizations by integrating with Open street map.
American Express, India
Entity Resolution image error
Developed an entity resolution algorithm to match business records using Support Vector classifier on features extracted using multiple string comparison functions.
Card Member Marketing image error
Identified card member lookalikes using k-Nearest Neighbor algorithm to create targeted merchant marketing campaigns that improved offer response rate by 65%.
Spend Trigger Model image error
Created a spend trigger model that predicts the likelihood of spend in a given industry using transaction patterns identified using association rule learning on past transaction data.