🛠 Tech Stack

💼 About This Role

You'll design self-supervised learning systems for Amazon's real-world robot fleet, turning terabytes of raw sensor data into structured training sets. Your work directly accelerates VLA foundation models and reinforcement learning for physical AI. This role offers direct impact on last-mile delivery robots operating globally.

🎯 What You'll Do

• Design and build scalable data pipelines for multi-modal sensor data.
• Develop self-supervised algorithms to extract features and pseudo-labels.
• Collaborate with VLA and RL teams to define data requirements and APIs.
• Architect evaluation strategies to track data and model quality.

📋 Requirements

• Master's in CS, ML, or Robotics or higher.
• 3+ years industry or research experience (PhD applicable).
• Expertise in self-supervised learning with multi-modal data.
• Proven experience with large-scale data pipelines (Spark, Kubeflow).
• Strong programming in Python and PyTorch.

✨ Nice to Have

• PhD in Robotics, Engineering, or related field.
• Publications at top-tier ML/robotics conferences (e.g., NeurIPS, CVPR).
• Experience with generative models for data augmentation.