Launchable applies machine learning to software testing
The start-up Launchable, which was co-founded by Kohsuke Kawaguchi, creator of the Jenkins CI/CD platform, is applying machine learning to software testing. The company’s technology predicts the likelihood of failure for each test given a change in the source code.
Still in stealth mode, Launchable is positioned to offer “smarter” testing and “faster” DevOps. The goal of the company’s technology is to eliminate slow feedback from tests, allowing users to run only the meaningful subset of tests in an order that minimises feedback delay.
Currently, most software projects run tests all the time, in no particular order, the Launchable website stresses. This can be wasteful when working on a small change in a large project. Developers know that only a small subset of tests are relevant, but there is no easy way to determine which tests those are.
The Launchable machine learning engine learns which tests are relevant by studying past changes and test results. Information from Git repos and test results from CI systems are refined into more meaningful data and then used to train the engine. The resulting prediction can be used in many ways, depending on where Launchable is deployed in the software development cycle. Launchable can be leveraged in intelligent integration tests, pull request validation, or the local development loop.
The company is seeking beta testers. Formerly the CTO at CloudBees, where he remains an advisor, Kawaguchi is co-CEO and co-founder of Launchable. The other co-CEO and co-founder is Harpreet Singh, who comes to Launchable from Atlassian and CloudBees. Both Kawaguchi and Singh also worked at Sun Microsystems.
IDG News Service