AI-Powered Document Processing & Mortgage Ecosystem

About the Project

Priority Financial Network (PFN) is a licensed mortgage lender operating across multiple U.S. states. 

Softices works with PFN on two fronts: 

• An AI-powered document processing platform that automatically classifies incoming mortgage documents, extracts and stores relevant data, and generates structured reports per document type.
• A suite of Ruby on Rails applications that power PFN's customer-facing websites, internal loan management systems, broker portals, and third-party user portals.

Together, these systems form the digital backbone of PFN's mortgage origination and processing workflow.

Business Needs

• Automatically classify uploaded mortgage documents and extract relevant data without manual review
• Generate structured, document-type-specific reports from processed data
• Generate structured import-ready files for a specific, high-priority document type (Tax Returns), per client requirements
• Capture additional data fields previously not extracted, and refine formatting of existing fields
• Provide raw extracted data for all other document types directly to the Ruby on Rails team for their own downstream processing
• Maintain and continuously improve a reliable, end-to-end document processing pipeline
• Provide customer-facing websites, internal loan management tools, broker portals, and third-party user portals under one connected mortgage ecosystem
• Integrate securely with third-party mortgage platforms (Encompass, Byte Software) for data exchange
• Automate loan workflows, lead generation, authentication, and document storage

The Challenge

• Classifying documents accurately based on a combination of image and text content, rather than either signal alone
• Extracting new, previously uncaptured data fields required for structured import files on Tax Return documents, while preserving accuracy on existing fields
• Transforming stored Tax Return extraction data into the strict structured format the client's import requirements demand, while keeping raw extracted data usable for the Ruby team across all other document types
• Giving the Ruby on Rails team visibility into AI model predictions during development, without a production-ready dashboard yet in place
• Securely integrating and synchronizing data in real time with external mortgage platforms (Encompass, Byte Software)
• Managing document storage, retrieval, and background processing at scale across multiple Rails applications
• Maintaining and evolving several interconnected Rails applications (customer sites, loan management, broker portals, third-party portals) as client requirements change

Solutions Offered

Results Achieved

• Ongoing improvements to document classification accuracy through combined image + text modeling
• Expanded data extraction coverage to support structured import file requirements for Tax Return documents
• A functioning internal dashboard giving the Rails team direct visibility into AI prediction quality
• A connected mortgage ecosystem of customer, broker, and third-party portals running on a shared backend architecture
• Real-time data synchronization with Encompass and Byte Software reducing manual data entry
• Scalable document storage and automated workflows supporting day-to-day loan processing

Technologies Used

Conclusion

By combining custom AI classification and extraction models with a connected suite of Ruby on Rails applications, Softices has helped Priority Financial Network automate document-heavy mortgage workflows while meeting the client's structured reporting requirements. The result is a scalable, evolving platform that supports PFN's borrowers, loan officers, brokers, and internal teams from a single, integrated ecosystem.