I am doing a survey on how long it takes and effort(monetary and skill) it takes to implement a research paper (AI/ML only)
I want to map in effort (money * hours) it takes just to implement, not apply or test on your application, just implement a paper, SOTA, like TurboQuant or LongBench with coding assistants, a team or how ever you work.
we have explored the idea a lot, and there is so many parts to it, scaling it to compute you have, resolving environments(if its an old paper), resolving bugs(introduced by coding assistants if used), validating if implementation is correct.
so if you can tell me the last research paper you've implemented and the workflow you used the journey, any way just I can map it to effort (money * hours).
I talked with an ML researcher, and he said with Claude a day at max for easier papers, but not sure about harder papers which large compute or infra. The main idea is to get an idea for them only.
And the workflow breakdown based on time, like most time spent on what, the exact SOTA code, environment resolving, scaling the SOTA to your compute or anything else, literally
We are building a platform to reproduce a research paper in just 10minutes so need inputs for that.
Really appreciate any help, thanking you in advance