Machine Learning Engineer - Video Generation Models

Apple; San Diego (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Apple is seeking a machine learning engineer to advance video generation models, with hands-on experience in training large generative models from pre-training to fine-tuning and efficient inference.

You will design training recipes, run large distributed training jobs across multi-node GPUs, and optimize both training and inference for production-scale deployment, collaborating with engineers and researchers across Apple.

Qualifications

• Bachelor's degree in Electrical Engineering, Computer Science, Computer Engineering, or related field with at least 3 years of relevant industry experience.
• Experience with large-scale generative model training for video generation.
• Experience running distributed training across multi-node GPU clusters.
• Strong Python programming skills and proficiency in PyTorch or JAX.

Responsibilities

• Pre-training video generation models, including architecture selection and training recipe design.
• Running, monitoring, and debugging large-scale distributed training jobs across multi-node GPU clusters.
• Improving training and inference efficiency through optimization techniques such as parallelism, mixed precision, checkpointing, and hardware utilization.
• Collaborating with product and research stakeholders to translate requirements into modeling and engineering tasks.

Skills

Python programming

Distributed training

Education

Bachelor's degree in EE/CS/CE

Tools

PyTorch

JAX

Job description

Summary

We are hiring a machine learning engineer with deep, hands-on experience training large generative models to help build our video generation models. You will work across pre-training, fine-tuning, and inference optimization, from designing the training recipe and running large distributed training jobs through making the resulting models efficient to run. As a member of the team, you will develop fundamental model capabilities and collaborate with engineers and researchers across Apple to advance our products.

Description

As a member of our fast-paced group, you'll have the unique and rewarding opportunity to shape upcoming products from Apple. We are looking for someone who has taken large generative models through the full lifecycle, from pre-training through fine-tuning and efficient inference, and can bring that depth to video, with the engineering skills to make that work reproducible and production-ready.

Key Responsibilities

• Pre-training video generation models, including architecture selection, training recipe design, hyperparameter and scaling decisions, and evaluation of model quality
• Running, monitoring, and debugging large-scale distributed training jobs, and diagnosing loss instabilities, divergence, and throughput regressions
• Improving training efficiency and cost, including parallelism strategy, mixed-precision training, checkpointing, and hardware utilization
• Improving inference efficiency through step distillation, few-step sampling, and quantization, and characterizing the resulting quality and latency trade-offs
• Adapting pre-trained models through fine-tuning, preference optimization, and knowledge distillation
• Building and maintaining the training and evaluation code the team depends on, with an emphasis on reproducibility and reliability
• Partnering with product and research stakeholders to translate requirements into modeling and engineering tasks

Minimum Qualifications

• Bachelor's degree in Electrical Engineering, Computer Science, Computer Engineering, or relevant degree, and a minimum of 3 years relevant industry experience
• Experience with large-scale generative model training for video generation
• Experience running distributed training across multi-node GPU clusters
• Strong software engineering skills in Python, with proficiency in a modern deep learning framework such as PyTorch or JAX

Preferred Qualifications

• MS or PhD in Electrical Engineering, Computer Science, or Computer Engineering
• Experience with video generation architectures, including diffusion or autoregressive models, temporal consistency, and long-horizon generation
• Experience contributing to major foundation or base model pre-training efforts, including scaling laws and transferring training recipes across model and training scales
• Experience with large-scale training operations, including parallelism strategies and diagnosing loss instability, divergence, or throughput regressions
• Experience improving and adapting trained models, such as step distillation, few-step sampling, or quantization for inference efficiency, and supervised fine-tuning, preference optimization, or knowledge distillation for quality
• Ability to work through ambiguity, collaborate across teams and disciplines, and communicate complex technical results clearly