Machine Learning Engineer, Apple Services Engineering

Apple Apple · Big Tech · San Francisco, CA +1 · Machine Learning and AI

Machine Learning Engineer at Apple Services GenAI & ML Frameworks team, focusing on bridging foundation model capabilities with real-world production systems. The role involves LLM continual pretraining, posttraining, agentic reinforcement learning, and agentic system optimization to improve LLM domain knowledge, tool use, reasoning, and system integration for user-facing features at scale.

What you'd actually do

  1. LLM training (domain-adaptive continual pretraining, post-training, preference optimization / RL such as GRPO-style methods)
  2. agentic systems (tool schemas, multi-turn reliability, rubric- or verifier-based learning loops)
  3. deployment-aware optimization (latency/cost/reliability tradeoffs, evaluation harnesses, and iterative improvement from production signals)

Skills

Required

  • Proficient programming skills in Python
  • Hands-on experience working with deep learning toolkits such as Jax, Tensorflow or PyTorch
  • Proven track record in training or deployment of large models or building large-scale distributed systems
  • Deep understanding of Deep Learning and Large Language Models (LLMs)
  • Natural Language Processing

Nice to have

  • PhD in a quantitative field, including Computer Science, Maths, Statistics, Physics, etc.
  • Experience building robust tooling around synthetic data generation, eval, and training pipelines for LLMs

What the JD emphasized

  • track record of turning LLM research into shipped capabilities
  • partner effectively with product, infra, and foundation model teams
  • lead ambiguous cross-LOB initiatives from problem definition through execution and scaling
  • building robust tooling around synthetic data generation, eval, and training pipelines for LLMs is strongly preferred
  • raise the bar on both research velocity and production readiness

Other signals

  • bridging foundation model capabilities with real-world production systems
  • improving LLM domain knowledge, tool use, reasoning, and system integration
  • bring cutting-edge models into user-facing features at scale