Multimodal ML Engineer: On-Device, Production Pipelines

Apple Inc.; Sunnyvale (CA); Hybride; USD 150 000 - 278 000; Plein temps

Résumé du poste

Apple Inc. in Sunnyvale is seeking a Machine Learning Engineer to build pipelines and production systems that turn multimodal foundation models into shipping Apple Intelligence features.

You will own end-to-end model delivery: building and scaling data curation and training pipelines, fine-tuning and optimizing large multimodal models for on-device and hybrid execution, and hardening approaches into robust, maintainable production systems under latency, memory, power, and privacy constraints.

Qualifications

• Experience in deep learning with multimodal systems (vision/language/etc).
• Proficiency in Python and in PyTorch or JAX.
• Experience with rapid prototyping, reproduction, and validation of research ideas.
• Ability to work in a collaborative environment.
• Ability to communicate results of analyses clearly.
• BS and a minimum of 3 years of relevant industry experience.

Responsabilités

• Build and scale data curation and training pipelines.
• Fine-tune and optimize large multimodal models for on-device and hybrid execution.
• Establish reproducible evaluation and regression testing for text and visual understanding.
• Collaborate with modeling, platform, hardware and product engineering teams across Apple.

Connaissances

Python

Deep learning

Multimodal systems

Collaborative work

Communication of analyses

Formation

BS in related field

Master/PhD preferred

Outils

PyTorch

JAX

Description du poste

Apple Inc. in Sunnyvale is seeking a Machine Learning Engineer to build pipelines and production systems that turn multimodal foundation models into shipping Apple Intelligence features.

You will own end-to-end model delivery: building and scaling data curation and training pipelines, fine-tuning and optimizing large multimodal models for on-device and hybrid execution, and hardening approaches into robust, maintainable production systems under latency, memory, power, and privacy constraints.