The API is built asynchronously based on RabbitMQ and allows integration with any payment gateways and other external systems.

Payments and credit charges are technically separated into different entities, reducing confusion when dealing with complex situations when external systems fail.

Accrual of interest and transfer to arrears is carried out by means of an overnight batch operation, which transfers the system to the next banking day. At the same time, payment by clients is made online - the system accepts payments, and immediately makes the procedure of allocation to accounts, changes the statuses of loans, and closes them. This makes the work of the loan operator more predictable for clients, and simplifies the logic of interaction between the engine and other systems.

The engine has a large set of methods to easily integrate with the loan pipeline - loan calculations, loan emulation, emulation of making payments, conditions on certain loans, etc.

The system provides transparent balance sheet accounting: all movements of funds on accounts are recorded using double entry transactional method.

In terms of accounting, the product provides for the maintenance of personal client accounts and company accounts, and the overall balance is maintained.

Product-engine allows to expand the set of credit products: the system architecture was developed with the expectation of future support for BNPL and Installment-products.

In addition, the engine maintains a log table that contains summarized information on each loan as of the last closed day. This data allows you to quickly and with minimal effort launch an in-depth analysis of the loan portfolio using external services.

The system is built using lightweight modern technologies (Kotlin, RabbitMQ, PostgreSQL) and is adapted to high loads out of the box. Without additional optimizations and tuning, a nightly batch operation to process 350 thousand loans and 4 million payments by execution time is within 1 hour.

The system provides for a flexible approach to correcting incorrect transactions that may come from external systems (incorrect payment dates, incorrect amounts, lack of reference information for splitting payments). The engine allows retrospective recalculation of the correct loan status for any corrections, while keeping the calculation base up to date at each point in time, preserving the integrity of any reporting forms. The correction is triggered by a special method and does not require manual intervention in the database.