A DevOps toolchain for data-driven decision making based on Machine Learning in the field of Corporate Performance Management

The thesis was carried out at a company that works with large financial institutions that have started to adopt machine learning to exploit their data. Currently the architecture of the system requires the use of a cloud provider but the client companies require this to be moved on premise. In order to bring the service on premise and with a view to future deployments of other services on cloud, it is necessary to identify a cloud-agnostic tool that allows both types of deployment; this will be chosen between the two main competitors currently on the market. Next, code has to be produced using the DevOps philosophy in order to automate the installation and management of the micro-services architecture; to do this, Infrastructure as Code and a Continuous Integration Continuous Deployment Pipeline will be used, again selecting the most suitable tool from the various available on the market. Finally, the savings in terms of time and money that a solution in full DevOps philosophy implies compared to a more classic solution will be measured. To do this, the BPMN notation will be used to run simulations within BIMP, a process simulator capable of providing the data necessary for comparison. The result shows that the savings are truly considerable and, for the installation alone, the saving is almost € 30,000.00, without taking into account the maintenance and updating phases, which will be fully automated and with minimal probability of failure