Document AI 📄

Best in class ML powered tool that solves industry document processing challenges.
On average, mortgage companies work with 
40 to 50 document types for a loan file.

That is a lot of document types to process. 
But with Document AI, it's a different story.
Document AI uses ML models to automate document data redaction, validation, and processing, significantly reducing manual effort. 

As an early design lead of the team (joining in Private Preview), I launched several core features from initial requirements to general availability.

Users

☺

Users ☺

Technical users

Engineers, data scientists and technical managers who want to create, monitor and maintain document processing models and workflows.

Labeling managers

Google specialists or internal subject matter experts who want to create and manage labeler pools & assign tasks to labelers if documents need reviewing.

Labelers

Manual Reviewers and human Labelers who want to help prepare training datasets and provide human review wherever required.

Fully automated UX

Users first choose from prebuilt processors or create custom processors for specific documents, upload data, use human-in-the-loop to validate any documents that don’t meet validation criteria then uptrain (manually or via HITL) processors to improve performance over time.

Bringing a new feature to an existing framework

I was faced with two challenges in this project:
  • Designing a brand new feature (HITL) with technical roadblock due to permission complications of bringing external labelers into the product as non technical user. 
  • Upon finding an external compatible tool to resolve the technical roadblock, aligning the look and feel of two separate tools that perform exact same tasks.

Labeling tool in Document AI that technical users use to manage dataset & processor training vs External labeling tool for human review labelers to complete HITL tasks

Human in the loop feature UI from Technical user perspective, setting requirements for labeling managers.

With my background expertise in visual & system design, I also redesigned the Home page and Processor gallery, including creation of original visual assets that went into production.