Create reproducible experiment logs from ad-hoc notebook runs

Your ad-hoc script runs leave gaps in experiment metadata and cause tracking drift across models. Centralized experiment records and consistent metadata let data scientists reproduce results before the next review cycle.