OCRed

OCRed automates ingestion, parsing, verification and insertion of toll records from scanned PDFs for a highway agency. The product handles multiple document templates, applies OCR and parsing rules, and presents a verification UI that pairs extracted tabular data with the original scans for human validation. The pipeline design emphasised fault-tolerant concurrency and idempotent processing to avoid duplication during retries.

Delivered capabilities included template-aware parsing, confidence scoring and retryable ETL steps. We recommended a machine-assisted review queue to prioritise low-confidence records, an audit API for downstream reconciliation, and workflow automation to trigger stakeholder notifications and downstream database syncs — all designed to reduce manual overhead and improve throughput.