AI Research Scientist, Multimodal Foundations

Apple, Inc.; Seattle (WA); On-site; USD 180,000 - 270,000; Full time

Job summary

Apple Inc. is seeking a research scientist to advance multimodal foundation models, focusing on architectures, pre-training objectives, and distillation techniques suitable for on-device deployment across Apple experiences.

The role spans model design, data strategies, optimization, and evaluation to balance capabilities with memory, latency, energy, and privacy constraints. You will explore teacher-student learning, distillation methods, and efficient architectures, aiming to transfer

Qualifications

• Hands-on experience designing, implementing, and running large-scale pre-training experiments for large language models.
• Proficiency with modern deep learning frameworks (PyTorch / JAX) and distributed training systems.
• Experience evaluating pre-trained models across language understanding, reasoning, instruction following, or multimodal capabilities.

Responsibilities

• Design and evaluate new multimodal foundation-model architectures, pre-training objectives, and data strategies.
• Develop and assess distillation techniques to transfer capabilities to smaller, on-device models.
• Co-design architectures and training pipelines for Apple silicon and on-device intelligence.
• Translate research ideas into foundation-model technologies applicable to Apple products and ecosystems.

Skills

LLM pre-training experiencia

PyTorch or JAX

Distributed training

Multimodal evaluation

Education

Master's degree in ML/CS

Job description

Apple Inc. is seeking a research scientist to advance multimodal foundation models, focusing on architectures, pre-training objectives, and distillation techniques suitable for on-device deployment across Apple experiences.

The role spans model design, data strategies, optimization, and evaluation to balance capabilities with memory, latency, energy, and privacy constraints. You will explore teacher-student learning, distillation methods, and efficient architectures, aiming to transfer