chore: rename training notebook to ramiru-decryptogen-train-task

docs: reframe README holdout claim as imputer robustness check

style: align README spelling and notation with the notebook

style: scrub AI-tell vocabulary and normalize notebook

style: make notebook warning suppression robust across worker processes

style: use seaborn vlag colormap for the correlation matrix

style: plain-markdown prose, softer correlation colormap, untruncated search table

docs: tighten Section 3.1 and 3.5 narrative

style: silence sklearn parallel warnings for clean notebook output

style: keep source comments plain-ASCII

docs: document the Section 3.5 hyperparameter search

feat: add hyperparameter search due-diligence (Section 3.5)

docs: document Yes/No casing normalization and holdout results

fix: normalize Yes/No casing at API boundary

fix: bump Docker base to python:3.12-slim so the pinned ML stack resolves

style: standardize status prints and keep the notebook ascii-clean

fix: pin ML stack to the versions that built the artifact

docs: add deployment and submission notes

feat: containerize API

feat: add FastAPI inference service

feat: fit final pipeline and export artifact

feat: add model interpretation

feat: compare models with stratified CV

feat: add preprocessing pipeline and baselines

feat: add EDA and data profile

chore: scaffold project