Tasks were planned and successfully completed
Machine Learning based application for documents processing
After 6 months from release, we usually ask customers about the financial results of a project to make sure we did our job well
Andersen team has spent more than 600 hours every week on the development of this solution
The application is able to process a document of 500 pages within 2 minutes, while a person would spend no less than a week.
Due to constant changes in legislation, investment funds and other financial market players incur tremendous expenses from monitoring these changes, in order to comply with all the required regulations. Lately, these kinds of expenses have increased by 45 times as it is normally based on ‘hand labor’. In order to find a software solution to this problem, the customer contacted us, and a team of experts from Andersen.
Machine Learning, Scala, C#, WC, WPF, MSSQL, ASP.NET, XAML
In order to build a product that would not only solve the task assigned but also be as convenient as possible, our FinTech specialists conducted a thorough research of the problem, studied the corresponding processes on the financial market in detail, analyzed users’ needs, and based on the gathered data, suggested several possible solutions to the challenge. After discussing them with the customer, it was decided to go with the concept that included AI and ML components.
Since the software includes complex functionality to solve the challenging task, it was necessary to build an effective interface so that the user doesn’t face any difficulties using the app. Therefore, the design is based on the principle of simplicity: a structured user-friendly system of widgets and tabs, simple icons and buttons, a well-thought combination of colors (shades of blue on the white background together with fresh colors).
The log-in system is simple and secure at the same time. The user simply enters their name and password, but the application launches only in case the administrator confirms the entering. The admin can also add a user to the database and provides the necessary rights.
The key feature of the application is an automatic analysis of bank documents with Artificial Intelligence and Machine Learning components. On the page below, the rules are created by taking data such as limits and selections from the database, calculating with complicated formulas, and adding to the corresponding table.
A module containing more than 50 widgets was created to enable automatic statistics gathering and evaluation. The process can be performed both automatically and manually. The module contains a lot of filters to ensure convenient work with data.
We developed a separate module aimed at storing documents in the program. Using AI, the system analyzes the documents added to the module and extracts rules from them for further implementation.
A module where investment units in the form of tables are created. Each investment unit can be connected to a certain jointure that will extract information from the given table.
Visualization of data from the database. Here, the user can select a table, view it, and configure the way they need choosing what is to be shown, which jointures it should be connected to, etc.
The database with all the rules. In the database, the user can search for the necessary data using various filters, as well as create a table right in here and then automatically fill it with data.
Our clients are provided with weekly reports enabling project processes monitoring and goals achievement progress.
Engineers with wide range of technical skills participated in this project. You can study their CVs and include in your team if need
The software solution developed by Andersen’s team automates processing of all incoming documents, analyzing pages at high speed. This is achieved with the help of Artificial Intelligence and Machine Learning components implemented in the application. As a result, the customer can save a big part of the budget which would otherwise have been used to run numerous departments. Besides, the application significantly declined the share of toxic assets.
Speed optimization rate
Toxic assets share
Virtual server hosting, container management, and serverless computing.
Understanding of the project
(Results based on 8 evaluation parameters)
(Results based on 7 evaluation parameters)
Tracking systems / task evaluation
(Results based on 9 evaluation parameters)
Code quallity control
(Results based on 9 evaluation parameters)
Auditing of QA (manual + auto)
(Results based on 14 evaluation parameters)
Total project score
Project growth measures the growth of the number of team members involved in the project and shows the difference in the number of team members at the end of the project comparing with its initiation
Project plan/fact matching measures how the project meets the initial budget and developing schedule.
67 system architects, lead, senior, full stack developers. Java is the core language we used for more than 25 Banking, HealthCare, and Insurance systems
Development of a complex high-functioning platform for the world’s largest online bank.