In 2021, Oleg Klimov, Vlad Guber, and Oleg Kiyashko embarked on a mission to develop Refact.ai, a platform designed to help companies adopt Generative AI (GenAI) for coding by offering enhanced customization and control for its users. Klimov and Kiyashko, who have nearly a decade of experience in AI-driven image recognition and security systems, collaborated on this initiative, while Guber, a childhood friend of Kiyashko from the South Ukrainian town of Yuzhnoukrainsk, joined the project.
“AI is poised to transform the very concept of engineering,” Klimov stated in an email interview. “As passionate software engineers, we felt compelled to prepare for this change by creating an independent software engineering system.” Indeed, many developers are recognizing the profound impact of AI in their field. A recent HackerRank survey revealed that 82% believe AI will fundamentally reshape the future of coding and software development.
Refact leverages generative AI models trained on permissively licensed code to operate its code-suggestions platform for developers. The 2023 survey from VC firm HeavyBit indicated that 63% of developers are already utilizing GenAI for coding tasks, embracing the transformation. Conversely, employers are more hesitant; a survey of C-suite executives and IT professionals revealed that 85% expressed concerns regarding GenAI’s privacy and security implications.
Major corporations, including Apple, Samsung, Goldman Sachs, Walmart, and Verizon, have restricted the internal use of GenAI tools out of fear of data breaches. So, what sets Refact apart? According to Klimov, it operates on-premises.
Similar to prominent GenAI coding tools like GitHub Copilot and Amazon CodeWhisperer, Refact can respond to natural language queries about code (for instance, “When was this dependency last updated?”), suggest code snippets, and adapt to enhance its performance with specific codebases. “You can think of it as a 'strong junior engineer' or an AI-powered colleague that needs a little guidance,” Klimov explained.
However, in a unique twist that differentiates it from most competitors, Refact does not require an internet connection and does not upload basic telemetry data, as Klimov claims. The platform can operate offline, on-premises, or via a managed cloud setup.
“We are committed to establishing robust controls and processes around data usage, security, and privacy,” Klimov said, acknowledging the challenges enterprises face while ensuring the confidentiality of customer information and fostering innovation.
Refact's platform is powered by compact, code-generating models trained on permissively licensed code, which Klimov identifies as a significant competitive advantage. Many code-generating tools trained on copyrighted or restrictively licensed code have been known to reproduce that code, creating potential liability risks for the companies utilizing them, according to some intellectual property experts. Vendors like GitHub and Amazon have implemented settings and policies to alleviate businesses’ concerns over these IP challenges, but success has been limited. A 2023 survey by Acrolinx revealed that almost a third of Fortune 500 companies cited intellectual property as their primary concern regarding generative AI.
“Our decision to use permissively licensed code for training was a direct response to our customers' demands,” Klimov noted. This privacy-aware and IP-conscious strategy has enabled Refact to secure $2 million in funding from undisclosed investors and initiate around 20 pilot projects with enterprise clients. Klimov asserts that the platform is generating revenue and is expected to reach “a few million” in annual earnings by this summer—an impressive feat given that established players like GitHub have struggled to turn a profit on their coding tools. Copilot, for instance, reportedly cost GitHub’s parent company Microsoft up to $80 per user per month due to the associated cloud processing costs.
Refact is part of a burgeoning sector of GenAI tools for programmers. The focus for the eight-person Refact team, based in London, is to upgrade the platform to run code autonomously, execute “multi-step” tasks, and self-test code. “We are actively developing a next-generation AI assistant that will debug the code it writes and function across any large codebase,” Klimov shared. “We are well-funded internally, with the necessary capital to keep enhancing our product. While we haven't enjoyed the bountiful funding typical of recent venture capital trends, we have benefitted from the enthusiasm of talented individuals eager to join the AI revolution and who see Refact as a platform where they can thrive and create lasting innovation.”
Ranking
