2025
Gikens OCR
Automated receipt and invoice OCR processing pipeline
Built and deployed a Linux-hosted OCR system for Gikens to process receipts and invoices quickly at scale. The pipeline runs HunyuanOCR three times per image to improve extraction reliability, then uses Gemini verification to resolve inconsistencies before structuring outputs into key accounting fields and CSV exports. Receipts with conflicting OCR passes are automatically flagged for manual review, enabling fast throughput while preserving data quality.
