Context
FEMA Public Assistance workflows involve large volumes of supporting documents that must be reviewed and classified accurately. Certificate review is a specific document intelligence problem where both text patterns and visual evidence can matter.
Challenge
Document classification could not rely on one method alone. Some certificates are identifiable through text, while others require visual layout or object-level recognition. The workflow needed to improve throughput and consistency while supporting QA/QC expectations in a public assistance environment.
My Role
I led the design and implementation of the AI certificate classification system for the VIEW platform. My work combined text matching logic with deep learning object detection to support automated identification and verification.
Approach
The system used text-based fuzzy matching to identify certificate language and patterns, while a deep learning object detection model provided visual classification support. The workflow was implemented in an AWS-based environment and aligned with operational review needs.
Output
The project produced an AI-enabled certificate classification workflow for the VIEW platform. It supported document identification, classification, and verification within FEMA PA CRC operations.
Impact
The system reduced manual review effort and improved review throughput and consistency. It demonstrated how text logic and computer vision can be combined for practical document intelligence in public-sector workflows.