Proceedings of the International scientific and practical conference ―Eastern European Forum of New Discoveries‖ (May 11-13, 2026) / Publisher website: www.naukainfo.com. – Kharkiv, Ukraine, 2026. - 208 p.

61 Keywords: cI/CD optimization, parallel execution, cross-platform automation, Agent management, pipelines, DevOps. The increasing demand for rapid software delivery has placed immense pressure on the performance of CI/CD pipelines [1], particularly in the context of cross- platform desktop automation. Unlike web applications that can be tested in standardized browser environments, desktop solutions for Windows and macOS require specialized build agents (runners) with specific operating system configurations. The primary challenge in such environments is the "sequential bottleneck," where the overall duration of the pipeline is dictated by the slowest test suite on the most resource-intensive platform. To overcome these limitations, a robust parallelization strategy must be implemented at the orchestration level. By decoupling test suites from individual machines and utilizing a distributed agent architecture, organizations can execute multiple E2E scenarios simultaneously. Within a C# .NET ecosystem, this is often achieved by integrating test runners with CI/CD platforms [2] like Azure DevOps [3], GitHub Actions [4], or Jenkins. Intelligent test sharding allows the workload to be balanced across available Windows and macOS agents, ensuring that hardware resources are utilized to their maximum capacity. This approach significantly reduces the time-to-feedback, allowing development teams to identify platform-specific regressions early in the deployment cycle. Effective agent management is another critical pillar of pipeline optimization. Managing "static" build agents often leads to environment drift and high maintenance overhead. The methodology proposed in this study advocates for the use of "ephemeral" or dynamic agents - virtual machines or containers that are provisioned on-demand and destroyed after the test run is complete. This ensures a clean, reproducible state for every execution, which is vital for the stability of Appium and WinAppDriver sessions. Furthermore, implementing tagging systems for agents allows the pipeline to intelligently route jobs: Windows-specific UI tests are