coreml

This skill provides comprehensive guidance for integrating Core ML models into iOS applications, enabling on-device machine learning inference without requiring network connectivity or cloud resources. It focuses exclusively on the Swift implementation side, covering model loading from various sources (.mlmodel, .mlpackage, .mlmodelc), configuration of compute resources, and execution of predictions using Apple's Core ML framework.

The skill encompasses critical integration patterns including auto-generated model classes for type-safe predictions, manual and asynchronous model loading strategies, compute unit configuration to optimize performance across CPU, GPU, and Neural Engine, MLTensor usage for modern iOS versions, MLMultiArray handling for older compatibility, and Vision framework integration for computer vision tasks. It also addresses multi-model pipelines, performance profiling techniques, and memory management strategies essential for production deployment.

Targeted at iOS developers building machine learning features into their apps, this skill is particularly valuable when implementing on-device inference for privacy-sensitive applications, offline-capable ML features, or performance-critical scenarios where latency matters. It supports iOS 26+ with Swift 6.3 while maintaining backward compatibility to iOS 14 where applicable, and explicitly excludes Python-side model conversion and optimization topics, which are covered by the complementary apple-on-device-ai skill.