Loading
A
ML Engineer, Apple Foundation Models
AppleCupertino, California, United States$150,400 - $277,600
Apply Join the team shaping the data foundation and intelligence for Apple's frontier foundation models. We believe that breakthrough AI capabilities are driven not only by model architecture and scale, but by the quality, diversity, and intelligence of the data used to train them. As part of the Apple Foundation Model team, you will help define how next-generation foundation models learn, reason, plan, and interact with the world, powering intelligent experiences used by billions of people. This is a rare opportunity to work at the intersection of cutting-edge AI research, large-scale training and data systems, and impactful consumer products.
As a member of Apple's Foundation Models team, you will develop the data strategies, pipelines, and methodologies that drive model capability across the full training lifecycle, including pre-training, mid-training, and post-training. You will work closely with researchers, engineers, and product teams to identify capability gaps, design data-centric solutions, and create high-quality training signals for reasoning, agentic behavior, multimodal understanding, tool use, and alignment. Your work may span large-scale data curation, synthetic data generation, data recipe development, model ablation, benchmark-driven optimization, reward modeling, evaluation systems, and data flywheels that continuously improve model performance. Every dataset, evaluation, and insight you contribute will directly influence the capabilities of the foundation models powering Apple's next generation of intelligent experiences.
- Drive data strategy and mixture design across the foundation model training lifecycle, including pre-training, mid-training, and post-training.
- Design and build scalable data generation, curation, and quality assessment systems for text, multimodal, reasoning, and agentic training data.
- Develop synthetic data pipelines that enable models to learn complex capabilities such as reasoning, planning, coding, tool use, and multimodal understanding.
- Create model self-improvement and self-iteration frameworks that leverage foundation models to generate, evaluate, refine, and evolve their own training data and behaviors.
- Pioneer data flywheels that transform model feedback, evaluations, and user interactions into high-quality training signals for continual capability advancement.
- Develop benchmark-driven methodologies to identify capability gaps, diagnose failure modes, and translate insights into targeted data interventions.
- Advance frontier capabilities in reasoning, agentic systems, alignment, and long-horizon task execution.
- Advance state-of-the-art techniques in data-centric AI, including reward modeling, preference learning, model self-evolution, and scalable alignment for foundation models.
- Demonstrated expertise in LLM or Multi-modal LLM with a publication record in relevant conferences (e.g., NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, KDD, ACL, ICASSP, InterSpeech) or a track record in applying deep learning techniques to products
- Proficient programming skills in Python and one of the deep learning toolkits such as JAX, PyTorch, or Tensorflow
- Ability to work in a collaborative environment
- Ph.D. in Computer Science, Machine Learning, Artificial Intelligence, or a related technical field, or equivalent practical experience.
- Experience developing data-centric solutions for foundation models, especially large-scale data flywheels.
- Experience improving foundation models using user interaction data, private data, or other real-world feedback signals while maintaining strong privacy and data governance standards.
- Experience building agentic systems, tool-use capabilities, and reasoning models.
- Experience with model self-improvement techniques.
- Experience developing or improving multimodal foundation models across text, vision, audio, and video.