Qualcomm Ventures, OurCrowd, and Samsung NEXT Ventures led a $15.5 million investment round for Tabnine, a business that creates a "AI-powered assistant" for software engineers, with participation from current investors Khosla Ventures and Headline Ventures. The funds will be used to improve the developer experience, introduce new features, and "strengthen" Tabnine's business offering, according to CEO Dror Weiss.


AI is increasingly being used to boost programming help. OpenAI's Codex, which drives GitHub's Copilot function, which gives ideas for lines of code inside development environments like Microsoft Visual Studio, is perhaps the finest example. According to the sales pitch, the tools would reduce overhead expenses while allowing coders to focus on more creative, less monotonous work.


Tabnine, which began as Codota in 2012, uses artificial intelligence to decipher code, autocompleting functions or "chunks" of code with a concept of their purpose and substance. The platform aims to learn individual best practices and warns of deviations using algorithms trained to grasp the semantic models of code.


In an email conversation with TechCrunch, Weiss said, "Tabnine... was formed by Eran Yahav and myself in 2017." According to him, the moniker "Tabnine" was inspired by a Waterloo-based firm of the same name that Codata bought in 2019. "Based on our past work on code analysis and simulation, we believed that, given the large amount of commonality and conventional patterns in code, AI would be an inevitability in the development process." We set out to create the AI code helper category and were the first to do so."


Tabnine uses small, "code-native" AI models trained from the ground up on specific programming languages or areas to provide suggestions on every keystroke as well as full line or function recommendations within integrated development environments like Android Studio, VSCode, IntelliJ, Webstorm, and Eclipse. Tabnine now offers more than a dozen popular language models, as well as "community" models trained by ecosystem partners, he added.



Tabnine's technique, according to Weiss, enables the platform to understand the "regularities" and patterns in code better than other code-generation systems — and to do it quickly. Weiss explained, "[Our models] provide clients the option to operate Tabnine on our cloud or on their network, as well as the capacity to train unique AI models that capture the exact patterns in their repositories." "Because inference cost and delay are substantially greater, [tools like] Copilot are confined to making ideas solely on new lines." Furthermore, they rely on a single massive monolithic AI model that can only be hosted by [big IT firms]."