Jenkins Creator Launches Startup To Speed Software Testing with Machine Learning

Kohsuke Kawaguchi, creator of the open source Jenkins continuous integration/continuous delivery (CI/CD) server, and Harpreet Singh, former head of the Bitbucket group at Atlassian, have launched a startup that's using machine learning (ML) to speed up the software testing process.

Their new company, Launchable, which emerged from stealth mode on Thursday, is developing a software-as-a-service (SaaS) product with the ability to predict the likelihood of a failure for each test case, given a change in the source code. The service will use ML to extract insights from the massive and growing amount of data generated by the increasingly automated software development process to make its predictions.

"As a developer, I've seen this problem of slow feedback from tests first-hand," Kawaguchi told ADTmag. "And as the guy who drove automation in the industry with Jenkins, it seemed to me that we could make use of all that data the automation is generating by applying machine learning to the problem. I thought we should be able to train the machine on the model and apply quantifiable metrics, instead of relying on human experience and gut instinct. We believe we can predict, with meaningful accuracy, what tests are more likely to catch a regression, given what has changed, and that translates to faster feedback to developers."

The strategy here is to run only a meaningful subset of tests, in the order that minimizes the feedback delay.

Kawaguchi (known as "KK") and Singh worked together at CloudBees, the chief commercial supporter of Jenkins. Singh left that company in 2018 to serve as GM of Atlassian's Bitbucket cloud group. Kawaguchi became an elite developer and architect at CloudBees, and he's been a part of the community throughout the evolution of this technology. His departure from the company was amicable: Its CEO and co-founder Sacha Labourey is an investor in the startup, and Kawaguchi will continue to be involved with the Jenkins community, he said.

Software testing has been a passion of Kawaguchi's since his days at Sun Microsystems, where he developed Jenkins as a fork of the Hudson CI server in 2011. Singh also worked at Sun and served as the first product manager for Hudson before working on Jenkins. They will serve as co-CEOs of the new company. They reportedly snagged $3.2 million in seed funding to get the ball rolling.

"KK and I got to talking about how the way we test now impacts developer productivity, and how machine learning could be used to address the problem," Singh said. "And then we started talking about doing a startup. We sat next to each other at CloudBees for eight years; it was an opportunity I couldn't pass up."

An ML engine is at the heart of the Launchable SaaS, but it's really all about the data, Singh said.

"We saw all these sales and marketing guys making data-driven decisions -- even more than the engineers, which was kind of embarrassing," Singh said. "So it became a mission for us to change that. It's kind of our north star."

The co-execs are currently talking with potential partners and recruiting engineers and data scientists. They offered no hard release date, but they said they expect a version of the Launchable SaaS to become generally available later this year. 

Posted by John K. Waters on January 23, 2020