Three sources, one truth. Across lakhs of rows.
A bank reconciliation platform that pairs transaction records across three independent systems and flags every mismatch.
- Role
- Full stack engineer
- Client
- UBI Bank
- At
- Oneture Technologies
- Scale
- Lakhs of rows
- Year
- 2025 — present
What reconciliation actually means here.
Every ATM transaction leaves a trace in three places: the switch that authorised it, the electronic journal on the machine, and the core banking system that moved the money.
When all three agree, nothing needs doing. When they do not, someone has to find the row and explain it, because each disagreement is somebody’s money.
The things that made it hard.
Binary machine files have to become structured data before anything can be matched.
Three feeds, three formats, all kept current by hand until now.
Operations work in Excel, so the result has to land in a sheet they can use.
How do you compare records that were never designed to be compared?
The three sources arrive in different shapes, and the machine journals are binary files that no one can read by eye.
2 passes at it.
Make the files readable
- Python scripts that turn the binary files into formatted, structured data.
One flow for all three sources
- Switch files, electronic journals and core banking data kept running through the same process.
- Formatted Excel output for operations, and the same data loaded into the database.
What came out of it.
Built to be reused: the same approach is being extended to more banks rather than rebuilt for each one.