For a software engineer I would rather recommend implementing from a reference software implementation. There are for example tons of "Model XX from scratch" in Python - which you can translate to Dotnet/Go.
Implementing code from a paper is almost its own skillset. Papers are often math heavy and they are information dense, with lots of references to exiting works. They are designed for communicating to other researchers in the same field/niche.
Implementing code from a paper is almost its own skillset. Papers are often math heavy and they are information dense, with lots of references to exiting works. They are designed for communicating to other researchers in the same field/niche.