Mr. Bhaskar Manchuri

Machine Learning–Assisted Energy-Aware Test Failure Detection

Abstract:

Automated testing in modern CI/CD pipelines incurs significant computational and energy costs due to the repeated execution of large test suites. This work presents a theory-driven framework for machine learning–assisted test failure detection that enables energy-aware test execution decisions. The approach models test selection as a constrained optimization problem that balances energy consumption, defect detection effectiveness, and operational constraints. Predictive models are used to estimate test failure likelihoods and guide selective execution rather than exhaustive testing. The formulation motivates approximation-based strategies that scale to large CI/CD environments. The proposed framework is independent of specific machine learning algorithms or application domains. Empirical evaluation on production-scale test workloads demonstrates substantial reductions in energy usage, while defect detection effectiveness is preserved or improved. The observed results align with the theoretical tradeoffs predicted by the model. Overall, this work shows that energy efficiency and software quality can be jointly optimized through principled, AI-assisted decision frameworks.  

Profile:

Bhaskar Manchuri is an industry researcher and Senior Manager of Software Engineering specializing in cloud engineering, distributed systems, performance engineering, and sustainable software practices, with applied work in AI/ML driven automation. He is recognized for expertise in cloud-native banking applications and test automation frameworks for enterprise-scale financial systems. He contributes to the research community through peer-reviewed conferences, invited reviewing, and conference leadership roles, and is an IEEE Senior Member and SCRS Fellow.