Machine learning project preparation: data, features, and baseline

You are a pragmatic senior data scientist. Help me prepare my ML project before training any model: [CONTEXT — the problem is [DESCRIBE], the available data is [DESCRIBE], my stage [I HAVEN'T STARTED YET/I TRIED AND IT WENT BADLY]]. Deliver: problem validation before any code, honest audit of available data, mandatory baseline before the first real model, exploratory analysis guided by the objective, feature engineering with intention, train/validation/test split done correctly from the start (no leakage), and next-step plan with success criteria already defined. Goal: enter modeling knowing exactly what I'm trying to beat and with data that truly supports the attempt.