shutterstock_564088018.jpg

Natural Language Processing and
Credit Scoring  

Challenge 

The VC-funded startup provides credit, savings, and other digital banking products to consumers who lack traditional banking accounts.

 

Their software team built the mobile application, but subcontracted the credit scoring algorithm that relied on alternative data points to evaluate a consumer’s credit worthiness to Gravitate. 

Solution  

Gravitate: 

  • Conducted a technical gap analysis and designed the data pipeline, database solution, NLP algorithms to process text, and predictive modeling for credit scoring.

  • To migrate from an R&D model to a production version, we wrote code using Airflow to streamline batch processing and a restful API for real-time scoring.

  •  Improved the natural language processing models.

  • To track patterns that could impact portfolio performance, we developed a reporting module.

Outcome   

A solution was delivered and operational in 3 months.