Self Service infrastructure
Our DevOps resources impact
Integrated ServiceNow-based self service forms to request infrastructure
Created API to manage and deploy terraform templates
Built fully automated business process for creating projects, groups, managing access and deploying infrastructure
The Challenge
Mobile and fixed network operators in Europe, managing thousands of projects in Google Cloud was facing an increasing maintenance burden of the requests to create and manage common types of workloads for Data Science projects. With multiple departments joining in the near future they decided to create a self serviced template for running and managing project templates across the organisation.
The Solution
The company has already been using the service now to automate some of the administrative processes and it has been chosen as a frontend for the automation effort.
Solid Potentials Engineers participate in cutting-edge design including access perimeters restricted with provider native tools as well as validating usability approaches of newly developed capabilities for newly developed GCP functionalities.
One of the main premises of the project was zero trust policy with multiple layers of isolation between development, staging and production. This required the ML solutions to be automatically built and delivered into target environments. To solve this, Solid Potential Engineers has built the IaC automation as well as internal project automation for provisioning, securing and orchestrating ML tools in project scope.
Moreover, there was an additional challenge for Data Scientists, who were not able to use dependencies directly, each of the dependencies of a deployed solution needed to be reviewed and vulnerability checked while still allowing hundreds of those solutions to be developed and iterated in real-time. We designed a gitops driven process that would evaluate the vulnerability of requested packages and accept or reject it into the system while keeping requesters informed of any potential issues.
Because of the project nature, Solid Potential Engineers needed to interface with multiple stakeholders, both technical and non technical. We developed unique strategy of documenting project boundaries by defining and building CLI to drive operations. This approach allowed us to hold API contracts with technical stakeholders, that we validated with a battery of unit tests. This approach allowed us to contain complexity into a well documented and tested code and as a side effect allowed us to augment stakeholder API capabilities by managing our own transformations. As a side effect, CLI allowed us to add system audits and dramatically simplify parts of orchestration and a finished system maintenance effort.
This was one of the most complicated projects for the customer yet, challenging not only technically but also from the management point of view as we were interfacing with stakeholders across company departments, external service providers and other companies building ML parts of the system at the same time. We have managed to deliver our part on time and within budget by keeping to our core strategy of data driven insight and keeping well documented interface between system boundaries.
