DeepCode, based in Zurich, claims that its system – essentially a tool for code analysis and improvement – is similar to Grammarly for programmers. The system uses a body of 250,000 rules, reads your public and private Github repositories, and shows you how to troubleshoot, stay compatible, and improve your programs.
Founded by Veselin Raychev, Counselor Martin Vechev, and Boris Paskalev, the team has extensive experience in machine learning and research on AI. This project is a derivative of the ETH in Switzerland and is an autonomous research project become programming utility.
How does it work? Pretty good. I ran one of my public filings through the system and I received 49 suggestions in 449 files. Patches range from literal code changes – changing name: String, to name: {type: String}, – to code suggestions that might be missing in function calls. This is an interesting tool, especially if you need help finding hidden bugs in your code. The advice given by this tool is also surprisingly accurate. Because he can build his own recommendations based on large amounts of code, he finds things that humans might miss.
"We have built a platform that includes the intent of the code," said Paskalev. "We autonomously understand millions of repositories and note the changes developers are making, then we train our AI engine with these changes and can provide unique suggestions for each line of code analyzed by our platform."
"Today, we have more than 250,000 rules and we grow every day," said Paskalev. "Our competition has to create rules manually and the biggest competitor has 3-4 000 rules and he has been working for years."
The company is self-financing and recently raised $ 1.1 million from Btov. The founders are serial entrepreneurs. Paskalev worked at Vistaprint and PPAG and Raychev worked for Google and is a researcher in the field of semantic programming machine learning.
More than a simple debugger, DeepCode "reads" and tries to compare the code to other implementations, offering you the best performance of each line. Now the team just needs to get programmers to use it.
"We have a single platform that includes software code in the same way that Grammarly understands the written language," Paskalev said. "This unique proposition allows us to save billions of dollars within the software development community with our first service and then be at the forefront of industry transformation towards a fully autonomous code synthesis. . "