Enhancing Lending Operational Efficiency with >99% Classification Accuracy
Banking & Financial ServicesBusiness Impacts
>99%
Page-Wise Accuracy
Reduced Manual Effort
>82%
Entity Extraction Text Accuracy
Customer Key Facts
- Location : North America
- Industry : Financial Services
Problem Context
The client is a leading Wholesale Lender in the United States, providing best-in-industry turn times, competitive rates, elite service, and technology. Their key offerings include home loans, refinance, mortgages, and other related services. They required an automated document processing environment to speed up their business processes to improve the long turnaround times and provide more transparency to customers on the status of the application.
Challenges
- To build a scalable and generalizable model that can classify 425 document types and extract entities from 34 document types
- To perform post-processing on Document AI API results that generalize across form types for Entity extraction
- To build a scalable API that can accommodate high traffic and volume of data
Technologies Used
Google Cloud Platform
Cloud Vision API
Document AI API
Google Bigquery
Python
Developed a Cloud-Native Document Processing AI Model
Solution
Quantiphi developed a GCP environment for users to upload various income and asset verification document types and receive a classification or entity extraction response using a template-based and template-free approach.
Quantiphi leveraged Cloud Vision API and GCP components to develop the automated pipeline which could be utilized to classify(on a page level) different financial documents such as tax forms with 98% accuracy.
Result
- Quantiphi delivered a cloud-native document processing AI model with a Page-Level Classification accuracy above 90% and the low API response time to deliver great business value
