Machine Learning Systems Engineer, Siri Agent Modeling

The Siri organization is looking for passionate Machine Learning Systems Engineers to join us in developing and shipping state-of-the-art generative AI technology to advance Siri and Apple Intelligence for Apple's customers. Siri is being elevated by the huge opportunities that AI brings.\\nThe organization is responsible for training on-device & cloud models, evaluating various approaches, pushing the envelope with the latest generative AI research developments, and ultimately delivering product critical models that power Siri and Apple Intelligence experiences. These models ship across a wide range of products at Apple, including iPhone, Mac, Apple Watch and more, enabling millions of people around the world to get things done every day. Our team provides an opportunity to be part of an incredible research and engineering organization at Apple. By joining the team, you will work with highly talented machine learning researchers and engineers, and work on meaningful, challenging and novel problems.

As a Machine Learning Systems Engineer, you will work closely with Siri modeling teams and other cross-functional teams to optimize model training and inference. You will be working across the ML stack at Apple, finding opportunities to make models performant, train quicker, and run faster on Apple's custom Apple Silicon. You will be joining a team that spans data, modeling, evaluation, deployment and working with engineers across ML infrastructure, inference, and framework teams. You will write production-level code to train and deploy models that will impact Apple's customers and enrich their lives. You are an ideal candidate if you: Are not afraid of CUDA OOM or NCCL errors

Experience in model lifecycle of training, evaluation, and deployment of models\nStrong understanding of Machine Learning (ML) model architectures (e.g. Transformers, CNN) and ML training loop\nStrong proficiency in Python and ML framework such as PyTorch\nBachelor's degree in Computer Science, Engineering, or related discipline, or equivalent industry/project experience

Collaborative with experience working in large inter-teams projects\nExpertise in ML and LLM optimization such as quantization, KV Cache, Speculative Decoding\nFamiliarity with ML training methodologies such as FSDP, DDP, and other parallelism\nExperience in an LLM training/eval library such as HuggingFace transformers, lm evaluation harness, Megatron-LM.\nExperience in optimizing LLM models and deploying LLM models\nProficiency in a compiled programming language (e.g. Swift, C/C++, Java)