On-device ML Infrastructure Engineer (ML User Experience APIs & Integration)

The On-Device Machine Learning team at Apple is responsible for enabling the Research to Production lifecycle of cutting edge machine learning models that power magical user experiences on Apple’s hardware and software platforms. Apple is the best place to do on-device machine learning, and this team sits at the heart of that discipline, interfacing with research, SW engineering, HW engineering, and products. The team builds critical infrastructure that begins with onboarding the latest machine learning architectures to embedded devices, optimization toolkits to optimize these models to better suit the target devices, machine learning compilers and runtimes to execute these models as efficiently as possible, and the benchmarking, analysis and debugging toolchain needed to improve on new model iterations. This infrastructure underpins most of Apple’s critical machine learning workflows across Camera, Siri, Health, Vision, etc., and as such is an integral part of Apple Intelligence. Our group is looking for an ML Infrastructure Engineer, with a focus on ML user experience APIs and Integration. The role is responsible for developing new ML model conversion & authoring APIs that will be a part of coremltools (CoreML’s authoring/conversion toolkit). The role is also responsible for integrating the APIs into internal and external systems (e.g., HuggingFace) to demonstrate the most efficient way of ingesting models into CoreML from these systems. This integration could involve a gamut of optimizations ranging from authored program optimizations (e.g., in PyTorch), to custom optimizations on CoreML’s model representation.

Description

As an engineer in this role, you will be primarily focused on developing and using APIs in coremltools to enable ML engineers to efficiently author/convert ML models to CoreML. You will be integrating coremltools into internal and external ML model repositories to evaluate and demonstrate how ML models can ingested into CoreML. You will ideate, design, and stress test the gamut of optimizations required to ingest these models, ranging from source level optimizations (e.g., in the PyTorch program), to custom optimizations after converting to CoreML’s model representation. The role requires a good understanding of ML modeling (architectures, training vs inference trade-offs, etc.), ML deployment optimizations (e.g., quantization), and a good understanding of designing Python APIs We are building the first end-to-end developer experience for ML development that, by taking advantage of Apple’s vertical integration, allows developers to iterate on model authoring, optimization, transformation, execution, debugging, profiling and analysis. The coremltools authoring and conversion APIs are the entrypoint to the rest of the infrastructure stack.

Minimum Qualifications

Bachelors in Computer Sciences, Engineering, or related discipline. Highly proficient in Python programming, familiarity with C++ is required. Proficiency in at least one ML authoring framework, such as PyTorch, TensorFlow, JAX, MLX. Strong understanding of ML fundamentals, including common architectures such as Transformers. Understanding of common ML inference optimizations, such as quantization, pruning, KV caching, etc.

Preferred Qualifications

Experience with any on-device ML stack, such as TFLite, ONNX, etc. Experience with designing Python APIs and production deployment of python packages is a strong plus. Experience with HuggingFace or any other model repository is a strong plus. Experience with MLIR/LLVM or any compiler toolchains is a strong plus. Good communication skills, including ability to communicate with cross-functional audiences.