Automated reviews for clean code — private beta launch
We are building a product that reviews code like how a developer would, for parameters that we’ve seen mostly matter in the real world, which is the ability to write clean code — code that others can read and understand.
AFAIK, there’s no such product out there. We were able to build it because we have 3 extremely senior developers (ex-ThoughtWorks) manually reviewing over a million lines of code for over 4 years, and collecting data systematically.
It would be fair to call Codu an experimental product. We’re at 95% accuracy right now, predicting for Readability, and working on OOPS usage prediction now. Also, we’ve currently built this for the recruitment space, where candidates solve a coding challenge as part of the hiring process. The current tools in the market (HackerRank and the like) mostly look for getting the testcases right, and how fast you do it — not necessarily what matters in the real world.
The reason we say we’ve built it for recruitment process is that we’ve used data from code submissions we got at Geektrust. So all the training data for our model is from candidate code submissions for a coding challenge. We’ve not tested it for smaller code snippets or large projects. We’ll cover that in the future.
If you’re looking to hire and want to filter candidates by quality of code, do check out Codu preview here.
While we’ve built it for candidate code submissions, you can just give their github URL and as long as the size of the repo is not too big, we will evaluate it. Write in to us and we’ll give free access to our private beta.
