Bank Statement OCR Processor Project notes.
A Go pipeline that reads bank statement PDFs and uses Gemini 2.5 Flash to extract structured transactions.

- Project
- Bank Statement OCR Processor
- Focus
- Document intelligence
- Explore
- Engineering archive
The problem.
Scanned statements, phone photos, and varying table layouts make transaction extraction difficult. The extracted rows also need to be checked against the balances in the source.
Tools & context
- Go
- Gemini 2.5 Flash
- OCR
- Data processing
How the workflow fits together.
- 01Built a concurrent Go ingestion pipeline to parse, pre-process, and batch upload document pages.
- 02Integrated Gemini 2.5 Flash via structured JSON outputs to perform OCR corrections and interpret transaction details.
- 03Added validation schemas to check running balances against the listed transactions.
Explore thecurrent project.
The pipeline produces structured transactions with balance checks for review.
Tell me what needs to work.
A new website, an existing codebase, or a workflow taking too much time. Start with the problem.
Start a project
