ART-Fintech ABS was tested in the AFT technological sandbox at Yadro server facilities

2024

Test at Yadro equipment stand

At a stand FinTech Associations (AFL) built on the company's equipment, YADRO the Russian development company "" microservice ABS ART Fintech tested the possibility of using the Russian infrastructure solutions in a typical financial organization. The AFL announced this on February 21, 2024.

ART-Fintech provided for testing the main functional modules of its microservice ABS: "Tariff Management System," "Conversion System," "Posting Generation System" and "Interest Calculation System." The process testings used only code that was automatically generated by designers and was not added by developers. Stress testing was also carried out, the purpose of which was to check the effectiveness of the system at the volumes of operations characteristic of large ones, banks for which a base of 100 million contracts was chosen. First, routine banking operations were tested separately, then - all at the same time when reproducing the procedure for closing a bank day, month or year. Testing was carried out at a special stand in the AFL technological sandbox implemented at servers VEGMAN and. storage system TATLIN.UNIFIED

Participants in the Russian financial market face a large-scale task of replacing foreign solutions with domestic ones to achieve technological sovereignty. In order to understand the capabilities of the solutions of Russian developers and determine the scope of their application, it is important to test them at special stands. The FinTech Association, together with YADRO, provided such an opportunity to market participants, which contributes to the rapid and effective implementation of domestic solutions, - said Oleg Morgun, head of the technology development department of the FinTech Association.

Testing of the ART-Fintech microservice ABS at the AFL stand built on YADRO equipment demonstrated excellent system speed and the ability to withstand even peak loads. And the closure of a month and a year from a large bank of the country would take a total of no more than a few hours when working on domestic decisions. The test results confirmed in practice the possibility of supporting the high-load operation of Russian banking systems built on freely distributed software components and launched on domestic hardware configurations, - said Larisa Stankevich, Doctor of Technical Sciences, Technical Director of ART-Fintech.

Testing applied cases of market participants at most levels of the technology stack of a typical financial institution allows you to better understand the possibilities and limitations of solutions, as well as significantly speed up their implementation in the domestic market. Russian banks are showing significant interest in automated banking systems (ABS) of domestic manufacturers. In this vein, the testing of microservice ABS by ART-Fintech at the AFL site can become a guide for other players and give an additional impetus to the development of the market, "said Pavel Karnauh, director of the YADRO sales technical support department.

Tests at the stand of the integrated solution by the Insurance House VSK

VSK Insurance House tested an integrated solution at the stand of the FinTech Association (AFL), built on YADRO equipment. The main purpose of the tests was to train and test neural networks and ML models that require high processor power. VSK Insurance House announced this on February 5, 2024.

At the first stage, training and testing of models of different architecture was carried out for the task of mapping (determining the correspondence between different data models) of text lists, in particular, predicting the belonging of text objects to classes of different levels. Models built using deep learning methods (Embedding + LSTM + Dense, implementation in Natasha, FastText and Keras/Pytorch-lightning and RuBERT) were evaluated by the ratio of training time to quality. It turned out that the learning accuracy when using graphics accelerators and conventional CPUs is comparable, but the learning speed on graphics accelerators has increased 15-80 times.

At the second stage of testing, models were trained and tested to solve the problems of improving entity recognition in unstructured text and photo recognition. During the tests, it was found that the use of graphics accelerators allows you to speed up the training process of small models up to 38 times (12-17M parameters). At the same time, the use of conventional CPUs is preferable for executing models in some cases.

For the tests, a stand was deployed at the FinTech Association based on YADRO equipment. The stand demonstrates the possibility of using Russian infrastructure solutions at most levels of the technological stack of a typical financial organization and includes VEGMAN servers, TATLIN.UNIFIED storage system, centralized SUPRIM management software.

Such a stand is an opportunity to conduct testing and determine the key parameters of the hardware and software complex required for the tasks based on domestic equipment for the planned lease or purchase. This significantly facilitated the budgeting process for 2024 for us in terms of hardware costs, - said the heads of the testing teams: head of the Fintech and Innovation Laboratory of the VSK Insurance House Alexandra Levushkina and head of the strategic project of the data analysis center, charging and risk technologies of the VSK Insurance House Alexey Chistyakov.

As part of the goal of achieving technological sovereignty, financial organizations are faced with the task of replacing the existing imported software and hardware complexes for the operation of core systems with domestic ones, without slowing down the development and introduction of innovations. The AFL stand allows you to successfully solve various problems, including in the field of ML: on document recognition, processing unstructured information flows, developing solutions for underwriting, etc., - said Oleg Morgun, head of the technology development department of the FinTech Association.

Insurance House VSK became the first user of our stand deployed in the AFL. We are grateful to our colleagues from VSK, with whom we managed to work on an extremely interesting task of using AI in the financial sphere. In 2024, we plan to focus on the study of reference architectures of complex solutions - primarily automated banking systems, to which almost all Russian banks are interested. At the same time, we will continue to conduct tests of solutions based on the Russian technological stack within the framework of various applied cases at the request of the Association members, - said Pavel Karnauh, director of the YADRO sales technical support department.