[narenst]

> As the first deep-learning-enabled product to launch on Github.com, this feature required careful design to ensure that the infrastructure would generalize to future projects.

It is surprising to see that this is the first time DL is run in production at GitHub. GitHub has a large amount of fairly structured data in the form of code, issues, etc. Plus they have been dabbling with DL for more than two years [1].

It could be that the business problems that are critical to the growth of GitHub product may not need DL. Solving problems like best-first-issues, code search are useful to the end user but may not effectively grow the business metrics.

[1] https://github.blog/2018-09-18-towards-natural-language-sema...