In this role, you will:

• Design and implement the end-to-end AI Flywheel, platforms for training, validation, deployment, and building a robust automated system.
• Build and maintain multi-node distributed training pipelines using tools like PyTorch DDP, Horovod, or Ray.
• Develop smart data mining and active learning strategies to prioritize valuable training data from petabyte-scale logs.
• Automate model evaluation and selection pipelines to support rapid iteration and closed-loop deployment.
• Build infrastructure for seamless model image packaging, validation, and rollout across Kodiak’s autonomous fleet and AI platform.
• Ensure that the flywheel is reliable, reproducible, and scalable, capable of learning from millions of real-world miles.

What you'll bring:

• Bachelor’s, Master’s, or PhD in Computer Science, Machine Learning, Robotics, or a related field.
• 3+ years of experience building production-grade ML infrastructure or model pipelines.
• Deep proficiency in Python and deep learning frameworks (e.g., PyTorch, TensorFlow).
• Experience with distributed training and pipeline orchestration (e.g., Airflow, Kubeflow, Dagster).
• Strong engineering fundamentals, debugging skills, and ability to scale systems.
• Passion for turning real-world data into self-improving AI systems.

Ideal candidate will also bring: 

• Experience in autonomous vehicles, robotics, or other sensor-rich real-world ML systems.
• Prior work with self-supervised learning, active learning, or large-scale data curation.
• Familiarity with containerization (Docker), model packaging, and deployment workflows.
• Comfort working in cross-functional teams with research scientists, infra engineers, and robotics experts.
• A mindset of ownership, experimentation, and systematic improvement.