AI Research Scientist: Multimodal Foundation Models - Architecture, Pre-Training & Distillation
Core
Designing, pre-training, and distilling multimodal foundation models for on-device deployment across the Apple ecosystem.
Role type
Senior IC AI Research Scientist (Multimodal Foundation Models)
Builds
Frontier multimodal foundation models and efficient distilled models for Apple silicon
Domain
AI Research, Multimodal Learning, On-device Intelligence
Deliverable
production ML models
Required skills
Large-scale LLM pre-training, Model architecture design, Deep learning frameworks (PyTorch/JAX), Distributed training systems, Transformer architectures, Knowledge distillation, Multimodal data strategies, Compute-optimal scaling, Teacher-student training pipelines
Preferred skills
Leading architecture experiments influencing large training runs, Research contributions in scaling laws/MoE/state-space models, Hardware-software co-design, Strong publication record, Open-source contributions
Technologies
PyTorch, JAX, Apple Silicon
Responsibilities
Develop and evaluate new model architectures and pre-training objectives, Design distillation methods preserving reasoning and multimodal capabilities under constrained capacity, Conduct experiments on compute-optimal scaling and data mixtures, Co-develop efficient models for on-device inference
Seniority
Senior, hands-on IC
