Digital Finance

The data basis consists of institutions’ regulatory XBRL-reports in the EBA framework as well as reports in accordance with national reporting requirements.

Leveraging their professional expertise, d-fine assembled a designated team of experts from the disciplines of regulatory reporting and applied AI.

Proofs of concept were developed in Python to identify patterns, outliers and trends through clustering, isolation forests and time-series forecasting based on the regulatory data submitted.

AI methods such as language models for semantic clustering of regulatory definitions were subject to thorough suitability tests.

To assess the risk development neural network models were applied to forecast regulatory reporting data based on macroeconomic indicators and market data.

The trustee acts as a neutral intermediary facilitating safe and dependable data exchange between parties. This allows for independent analyses of combined data without data providers losing data sovereignty.

As consortium coordinator d-fine is responsible for the  overall project management. This includes planning and coordinating the cooperation between consortium members.

d-fine develops the trustee’s technical platform as a microservice application in Kubernetes, with data contracting and data transfer through the Eclipse Dataspace Connector.

The consortium is involved in shaping the European legal framework for data trustees through continuous dialogue with legislators and regulators and prepares EuroDaT’s integration into this framework in due course.

d-fine developed several Proof-of-Concepts (PoCs) to evaluate different technologies for secure custody of crypto assets (e.g. HSM and MPC based solutions).

This was complemented by the development of a PoC approach for money laundering compliance.

d-fine assisted in designing the target operating model of a corresponding custody, trading and anti-money laundering solution.

In addition, the impact of the MiCA proposal on the future business model was analysed and compared with the current regulatory requirements.

d-fine analyzed existing offerings and technologies in the market, including alternative solutions to SSI and their security.

d-fine identified critical security gaps that need to be addressed for a secure offering of SSI-based services. Ways to close these gaps were developed and evaluated for their feasibility from the client's perspective.

The selected solution was conceptually designed in detail and a target operating model was created, intended as the basis for a patent.

To illustrate the developed process flow, d-fine designed a demonstrator.

d-fine revised the internal information security requirements and methods based on the international ISO standards 27001 ff. The information security requirements were thereby classified in a structured and easily comprehensible document hierarchy.

The risk methodology was based on a protection requirements analysis of all components of the information network. Inheritance between interdependent information network components was taken into account.

The challenge of the project was to design a methodology that was both easy to use by information owners and met operational risk management requirements for risk assessment and risk acceptance.

To explain a complex AI model, it must be transformed into a simpler representation. The challenge is to find a good compromise between the consideration of the properties of the AI model and the simplicity of the representation as well as a suitable approach.

Central tasks were the conception and design of an XAI (eXplainability AI) module for complex AI model ensembles using model-agnostic methods (LIME, SHAP and ANCHOR), including the evaluation of the technical-mathematical foundation of the XAI methods and their suitability for the use case, as well as the evaluation of alternatives (e.g. model-specific methods) and the weighing of advantages and disadvantages.

The Explainability Module was integrated into the customer's own infrastructure and a reporting solution with suitable visualizations of the results was built.

As general contractor, d-fine implemented a solution consisting of a web portal for the supervised companies, a data science platform for massive data analysis and web applications for the financial supervisors.

The platform runs on over 100 servers and uses a process engine to uniformly map standard and exception processes.

The technical basis was a mixture of standard technologies such as Camunda or Tableau, which were integrated with in-house developments for web applications, back ends and a data science platform.

The first release went into production after just nine months, with 70 d-fine consultants involved in implementation at peak times.

Several stress tests and other supervisory exercises have been successfully performed on the platform since go-live.

d-fine makes a significant contribution in all areas of the project - from the regulatory-compliant requirements specification to the implementation of data migration and project management.

In this project, we are convincing through the combination of our professional expertise in the areas of payment transactions, compliance and process digitization as well as our technical skills in system integration.

The result is, on the one hand, a scalable and modern IT infrastructure via which payment orders can be processed efficiently and in compliance with regulatory requirements, and, on the other hand, a state-of-the-art e-banking solution for the customers.

In a first step, d-fine contributed significantly to standardizing the supplier files, which had previously been created manually, and to automating their technical processing.

This enabled the automation of the infrastructure around the database and reporting at group level in parallel with the design and coordination of new technical interfaces.

The result was quick wins after each project phase, so that the customer could benefit from digitization as quickly as possible.