Developers are hard to find, hire, and keep. Great developers are extremely hard to find, hire, and keep. Thus it’s hard to scale software development teams horizontally (increase count) and almost impossible to scale them vertically.
We are going to create AI agent that would serve a software developer role. It will communicate with its human peers via chat (using restricted English), tracking system comments, and code.
Known competitors approach the task either from pure Deep Learning perspective (end-to-end learning), or from neurophysiological/NLP/knowledge management point of view.
First approach is mere brute force and requires huge computational resources and datasets. Generated solutions would likely be non-interpretable. Second approach is non-differentiable, thus requiring lots of manual labour to maintain and update.
On the other hands, founder has 18 year experience of managing distributed software developers, that require special attention due to their need to be the best or to the things in a different way. Due to daily root cause analysis and AARs, founders developed unique and detailed software development process, that could be transferred to existing deep learning methods.
Public proof of concept to be released during 2018q3