Idea
Machine learning platform for software teams to select targeted tests, reducing CI time and maintaining high fault detection.
Research Paper
Core Innovation
This paper introduces T-TS, a novel test selection approach that models commits as Bags-of-Words of changed files and integrates cross-file and predictive features without relying on coverage maps. This enables efficient and scalable test selection that significantly reduces execution time while preserving fault detection effectiveness compared to existing methods.
Market Size (TAM)
$2–10B TAM, $1–2B SAM; assumption: large global software development market with growing CI/CD adoption and test automation needs.
Potential Customers & Pain Points
- Software Development Teams Facing Long CI Times
- DevOps Teams Needing Faster Feedback Loops
- Enterprises Managing Large Codebases and Test Suites
Business Model
SaaS subscription model offering integration APIs and enterprise support for continuous integration platforms.
Competitive Landscape
- Google Bazel
- Microsoft Azure DevOps
- GitLab CI
Implementation Challenges
- Integration with diverse CI/CD pipelines
- Adoption resistance due to existing workflows
- Data privacy concerns with commit and test data
Validation Strategy
- Pilot deployment with select enterprise software teams
- Benchmark against existing test selection tools on live CI data
- Collect user feedback and iterate on feature improvements
Research Paper Overview
Targeted Test Selection Approach in Continuous Integration
Summary
This paper proposes Targeted Test Selection (T-TS), a machine learning method for efficient test selection in continuous integration. It represents commits as Bags-of-Words of changed files, includes cross-file and predictive features, and avoids coverage maps. Deployed in production, T-TS significantly reduces test execution time and pipeline duration while maintaining high fault detection rates.