New to this field need some guidance with my project( marine reasoning) [D]

learnt about vector space , fields , and applications whatever i could then went on with learning python libraries like numpy , scikit learn , pandas , also had a bit of knowledge about tensorflow and how to use pytorch but now i stand so clueless when i try to apply my knowledge in my first project.

So idk why but i decided somehow that i wanna make a project in "marine reasoning " and i feel clueless about is it just supervised learning reinforcement learning or mix of both and how much data should i acquire or how do i cleanse my data is the data even enough what should be my pipeline what are the steps i dont have any guidance.

I just wish someone who has keen knowledge in multi stage framework LLMs can give me the proper PIPELINE for the model and where should i start with , with proper links and tools also any inputs from anyone's side would be highly appreciated . ALSO IF ANYONE CAN GUIDE ME THRU ALL THE STEPS OR WHAT I NEED TO LEARN MORE .