The time when we knew what we as an IT department had to deliver in a year’s time is fortunately behind us. We only deliver one product: flexibility. The best way of doing this is through agile working in DevOps teams and by continuously delivering new features to our business units. I’m currently working in the Cloud Integration department, where I’m responsible for developing Azure Cloud capabilities that meet the requirements of NN Group.


To be able to deliver quickly enough, we need to ensure that no more operations work is needed on the services delivered. And to maintain momentum during the development stage, we have to ensure that all releases can be built, tested and rolled out automatically. By working safely, we can continuously test and roll out. We call this Continuous Integration/Continuous Development (CI/CD). Application developers have been working on this for a while now, but it’s a new area for Infrastructure. At NN Group, we are working on building pipelines using the tooling available in the open source market. This means that we are able to roll out new services in the Public Cloud. This is really challenging, not least because we encounter issues that need our immediate attention.

Azure Machine Learning

One such issue recently arose in relation to Azure Machine Learning. This Microsoft cloud service allows you to automatically discover models in large quantities of data. These models are generally so complex that they can’t be discovered manually. We can make predictions based on the models found.

We need to be constantly aware that our work doesn’t contravene our privacy and security policy. This is why we only grant authorisations for Azure based on roles that are linked to groups and users. However, there are no standard roles available in Azure that we can use for the authorisations required, so that you have access to the sources needed. This is why we have created new roles for the Data Tech Team for the data analysts.

Cost efficiency as well as speed

Because our IT rollout is highly automated, we can deliver new IT environments to our business units quickly. Our aim is to ensure that a customer can order the environment using self-service and that a new environment can be prepared without manual interventions, enabling the customer to get to work straight away. Once the environment has been rolled out, our services automatically scale up and down, meaning that there is no unnecessary IT expenditure. Given that we also have to deliver the security of the environment at the same time, this is still a significant challenge!

Fail fast, learn and adapt

Not churning out plans ad infinitum is the only way to continue improving and to allow us to carry on developing more services faster and faster. We put into practice an idea that appears viable and useful. If we don’t succeed, or if it later transpires to be not such a good idea, at least we’ve learned something and found some new ideas. To be able to develop new services, we make sure that we are and remain up-to-date with all developments in the Public Cloud and the tooling that can be used to roll out and manage infrastructure automatically. This makes our DevOps team a very challenging place to work!

Koen van Amerongen
DevOps Engineer Cloud Infrastructure