All case studies
Financial services
Automated financial reconciliation with bank statement OCR
Extracting data from PDF statements across different banks with OCR, plus reconciliation rules integrated into a legacy financial platform.
- Automation
- Legacy integration
- Custom development
Context
Reconciling bank statements against entries in a legacy financial platform was largely manual. Each bank delivered statements in a different format — some as digital PDFs, others scanned — and finding discrepancies took up the team’s time.
Challenge
- Reliably extract transactions from bank-specific formats, including scanned documents.
- Integrate with the existing Sybase platform without replacing it.
- Support period closing by flagging discrepancies for investigation.
Approach
I built a solution in Java with Spring Boot:
- PDF extraction flows for digital documents and Tesseract OCR for scanned ones.
- Bank-specific parsers that isolate each layout’s quirks, so new banks can be added without touching the rest.
- Reconciliation rules that match statements against platform entries and surface discrepancies.
- Period-close controls, so the team knows what has been reconciled and what still needs investigation.
Outcome
- Automated reconciliation, with the team focusing only on discrepancies.
- The legacy platform remained the system of record, now fed with structured data.
- An extensible base for new banks and statement formats.