Photography: Ben Itzhaki
Software companies have moved to rely on CI/CD today, where every small change in the code activates a slow and complex system of automatic tests that make sure that the change is correct, and does not harm the user experience. As a result, the pace of the CI/CD system dictates the delivery pace of companies – which becomes one of the biggest barriers that prevent organizations from innovating and significantly harms the development experience in the organization. A new Israeli startup has a way to save some of this headache, and above all to improve the pace of development.
inject some ML into your CI/CD
The Israeli redefine.dev has developed a platform that connects to the CI/CD systems of organizations, and uses ML models to automatically identify which parts of your code have changed, and as a result, run the appropriate tests dynamically. “We collect all the information about the changes made to the code and track all the failed tests and this is how we train a model to automatically identify the relationships that indicate the relevant tests over time. With the help of code analysis, we process each change into basic characteristics that can be entered as inputs to a model that learns to prioritize the tests According to the order of their relevance to the change that happened in the code,” he explains In a conversation with Gigtime The CEO of the company and one of the founders – Lior Neuman. He explains that the model is trained on information from the customers along with open source and since the use of ML for CI/CD optimization is relatively new – they developed many elements from scratch, including the definition of the model, process Training and information gathering.
The development of Redifine tries to improve the existing situation, in which companies often invest More resources in more powerful machines and increasing the amount of machines working at the same time to shorten the CI time – only without all the excess costs. According to the company, Adding machines or investing heavily in speed does not necessarily produce a significant improvement in the quality of development in organizations. “The ability of the platform allows us to give feedback 10 times faster than the existing systems and still maintain the same quality level of code,” Neuman added. He emphasizes that the model at the heart of their platform is an open source model, to which in-house developments have been added: “For specific needs, we use OpenAI models, for example, to produce an automatic correction when a test fails.”
Only in the last few weeks, an Israeli startup was revealed that developed a product designed to assist development and based on ML, and a day after its recruitment announcement, Microsoft and GitHub presented a similar feature. What guarantees that your product is not a feature that will be swallowed in a second by one of the giants?
“Unlike products that focus on creating code, the problem we solve cannot be solved directly by Generative AI. In the coming years, we will see how AI writes much more code than humans, and the barrier of the amount of manpower on the way to innovation is decreasing. We and our investors see Redfine Part of this change, we enable companies today to leverage the capabilities of ML to improve their performance. In the near future with millions of new lines of code written by AI, organizations will also have to adapt the supporting infrastructures for example CI/CD to the new world.”
Recruiting from Apsagon founders and a former executive at Twitch
Today (Monday) redefine was revealed with the completion of a seed round raising in the amount of 8.5 million dollars. The round was led by Gillot Capital and a number of private investors participated, including Nitzan Shapira and Ran Rabenzaft, founders of Gonaps; Efrat Rapoport, director of Salesforce Israel; Assaf Ezra founder of Granolite; Simon Freeman, formerly Twitch’s VP of Development; and Farhan Tawar, Shopify’s VP of Development. The company was founded in 2022 by Neuman and Eyal Regev (CTO) and employs about 25 people in the development center in Tel Aviv. Neumann tells us that the company already has paying customers today.