The Asset management team at a.s.r. was facing significant fragmentation between development and production environments. In the existing setup, quantitative tools and models (ranging from traditional calculation scripts to advanced machine learning assets) shared the same infrastructure, making production workloads vulnerable to errors stemming from ongoing development activity.

This lack of separation heightened operational risk, one developer’s changes could break production models, or inconsistent deployments might create confusion between teams. Moreover, as a.s.r. acquired more entities such as Aegon and their workloads, the challenge grew: more teams, more models, and an increased need for standardized processes and security.

The goals for the project were clear but ambitious:

• Establish distinct development, acceptance, and production environments (an OTAP street).
• Migrate key legacy tools (written in PHP, R, and Python) to a modern Python-based workflow.
• Ensure security and governance are integrated from the ground up.
• Enable teams to onboard and leverage the platform with minimal friction.
• Create an environment where both existing and new models could be developed, tested, and pushed to production with confidence.

From the outset, the new platform would be “Azure first,” aligning with a.s.r.’s broader IT strategy. Xomnia embedded data engineers within the capital management unit to deliver a solution closely integrated with both technical and organizational realities.
Key components of the solution included:

• Azure Data Factory for all data ingestion needs, from on-premises databases to web scraping, ensuring consistent, automated data pipelines.
• Azure Data Lake for scalable, secure storage of all input, intermediate, and output data.
• Azure Synapse and Azure ML to handle data transformations and model orchestration, supporting both classical quantitative models and machine learning workflows.
• Azure DevOps as the backbone for code management and continuous integration/continuous deployment (CI/CD), ensuring all code changes were properly tracked, reviewed, and safely deployed across the different environments.
• Implementation of the “OTAP” (development, acceptance, test, production) street, providing clear boundaries between experimental work and core production processes.
• Migration of legacy quantitative tools to standardized, Python-based implementations, and adoption of best practices for reproducibility and governance.
• Secure connections to on-premises resources, and integration with company-wide identity, access management, and data governance requirements.

The project unfolded over nearly three years, with quarterly planning cycles enabling iterative expansion. The platform started with a handful of models and two or three engineers, but quickly scaled to over a hundred models and more than sixty active users within capital management alone. Regular onboarding sessions helped different teams migrate their workloads and adapt to cloud workflows using Azure’s resources and best practices.

The implementation of the data platform has had a broad impact across a.s.r.’s asset management operations:

• Efficiency: Teams are able to develop, test, and deploy new models or update existing ones much more easily, leading to a surge in productivity and enthusiasm once initial adoption hurdles were overcome.
• Security and governance: The clear separation between development and production means that production workloads are not disrupted by in-progress development, and security requirements are consistently met across all environments.
• Scalability: The platform now supports over a hundred models, accommodating both new and migrated workloads. As more teams onboard, they benefit from established architecture and deployment pipelines, reducing the need to “reinvent the wheel.”
• Collaboration: The self-service design empowers individual teams to manage their own pipelines and models, while maintaining consistency with the broader platform standards. Onboarding efforts are ongoing, ensuring that more units across a.s.r. can participate and innovate without unnecessary barriers.

Although assigning specific numbers to efficiency gains is challenging, the new environment has made it noticeably easier for data and quantitative professionals to focus on their core tasks. Security, reproducibility, and operational consistency have all improved, reducing risk and freeing up resources for further innovation.

Looking forward, the platform is positioned for continued evolution. a.s.r. plans to onboard more teams and projects, further expanding the reach of the data platform and modern way of working. As the organization explores new technologies, the groundwork provided by a unified, scalable data environment ensures future developments can build on a strong foundation.