I built a ML training framework from scratch in cuda in 2016, for earthquake forecasting - I know quite well how to solve low level kernel optimization problems :)
I'm also regularly a rust developer and I do know REGEX thanks.
Building that project today, would not be impressive or particularly hard. There are great references available today to illustrate how to build that. Sure LLMs aren't 100% accurate, neither is an intern or junior dev you might offload a task to. Part of the job of a software engineer is collaboration. Understanding how to collaborate with teammates (local, AI, outsourced) and how to ship when your teammates make mistakes is quite important to know how to resolve.
That's why CI/CD and TTD exists! If you have a project where correctness is critical, why haven't you written a test to validate that correctness?
Comments
I built a ML training framework from scratch in cuda in 2016, for earthquake forecasting - I know quite well how to solve low level kernel optimization problems :) I'm also regularly a rust developer and I do know REGEX thanks.
Building that project today, would not be impressive or particularly hard. There are great references available today to illustrate how to build that. Sure LLMs aren't 100% accurate, neither is an intern or junior dev you might offload a task to. Part of the job of a software engineer is collaboration. Understanding how to collaborate with teammates (local, AI, outsourced) and how to ship when your teammates make mistakes is quite important to know how to resolve.
That's why CI/CD and TTD exists! If you have a project where correctness is critical, why haven't you written a test to validate that correctness?