Build trustworthy ML pipelines for large-scale scientific data analysis with calibrated uncertainties, simulation-based inference, and physics-informed constraints. Use when: 'build a photometric redshift pipeline', 'add uncertainty quantification to my ML model', 'create a simulation-based inference workflow', 'integrate ML into a scientific analysis pipeline', 'validate ML predictions against simulations', 'set up active learning for survey data'.