The typical differences between one implementation and another is: Both implementations might work for the original use-case in the happy-path case. However as

The typical differences between one implementation and another is: Both implementations might work for the original use-case in the happy-path case. However as soon as the first unexpected error occurs, the bad implementation might break down. It might be a lot bigger in size than the better one and contain a lot of duplication. And it might not be extensible or maintainable, and have not (regression) tests. And of course the time it took to produce the codebases might have been very different. Those are typical differences one sees between code that had been written at more professional software shops, and code that had been written elsewhere (e.g. by electrical engineers, students, non-engineers, etc.). However I have no idea how things look like at Google or Infosys, so I can't speak for those.