Document recognition for corporate lending

Speed up the review of corporate loan applications. Each application is a package of documents of different formats and quality that requires careful checking.

A system was needed that reliably extracts details, amounts and figures from scans and photos.

One of the largest banks in Russia. Name under NDA.

We developed a primary document recognition system for the bank's credit process.

1. 01Classification of incoming documents by type
2. 02Recognition of acts, invoices, charters, contracts, accounting and financial statements, passports
3. 03Extraction of key fields and transfer of structured data into the credit process
4. 04Recognition quality control on difficult documents

In 2022–2023 the system showed the best recognition quality among the bank's contractors. Data from documents enters the credit process without manual entry.