Engineering patterns for ML/DL research and LLM training infrastructure. Use this skill when: (1) Building training pipelines (SFT, RLHF, DPO, etc.), (2) Creating ML experiment infrastructure, (3) Designing distributed training systems, (4) Research projects requiring reproducibility, (5) Any deep learning project beyond simple scripts. Complements engineering-code skill with ML-specific patterns: config-driven architecture, component registry, experiment management, and stable/experimental layer separation.