News

Kite Beefs Up Its AI-Powered Code Assistant for Teams

• By John K. Waters
• 02/03/2021

This week the team behind the Kite code completion tool added a self-hosted machine learning (ML) engine to its product roster. Designed to provide personalized code completion at scale, the new Kite Team Server supplies enterprise development teams with personalized code completion in 16 programming languages across 16 editors and IDEs.

The San Francisco-based company's Kite, which suggests code snippets for developers in real time, is billed as "an AI copilot that helps programmers write code." The AI-powered coding assistant helps programmers code faster with multi-line completions.

"Programmers spend too much time on repetitive tasks, such as looking up documentation on the Web, fixing simple errors, and writing boilerplate code," the company said in a statement. "These inefficiencies make programming less enjoyable, block creative problem solving, and contribute to teams shipping less software. By surfacing the right information at the right time, Kite helps programmers write better code faster."

The Kite Team Server provides better code completions, more often, thanks to the deployment of an advanced GPU that sits inside a company's internal network. The Kite Team Server GPU trains personalized ML models based on a company's proprietary codebase, so that code never leaves a company's network. Completions shown by Kite Team Server guide developers to repeat idioms and patterns from internal code. 

Compared to Kite Free, the company's flagship offering for individuals, the average completion length doubles from two words to four words in the Team Server, the company says, and completions used per line of code written increases by 40%. These code suggestions help developers stay in flow and write code faster.
Kite Team Server provides completions in all of the most popular programming languages and IDEs.

Kite Team Server includes a number of new features, including:

• Larger machine learning models: Kite Team Server is powered by server-scale ML models that use 25X as many parameters as the ones used for Kite Free. Completions powered by Kite Team Server are longer and smarter. 

• Custom models trained on company code: Kite Team Server allows companies to securely train machine learning models on their proprietary code base. Personalized completions are more relevant and guide developers to repeat idioms and patterns from internal code. Junior developers can more easily navigate unfamiliar or undocumented code. 

• Enterprise grade security.

"Over 400,000 developers trust Kite," said Adam Smith, CEO and co-founder, in a statement, "including software engineers at 35% of the Fortune 500, and so this technology has been stress tested the past few years," We have grown from a single Python code completion tool into present-day where enterprise teams can use Kite across 16 programming languages. Our users asked for this enterprise version of Kite. So this completions engine is a solution that is agile enough to provide value to teams of just a few developers and robust enough to support thousands of simultaneous users. Our new enterprise-grade application architecture satisfies the most rigorous IT and security requirements."