AutoFACS CV/ML is the computer-vision and machine-learning track of AutoFACS: a FACS-informed facial expression recognition project built around staged classification, curation-driven data improvement, and reliability-aware evaluation under Apple Silicon / MPS constraints. This public repository focuses on architecture, methods, selected V40 results, and the project's engineering evolution while leaving out private datasets, checkpoints, and other disclosure-sensitive internals.