Senior RL Engineer for Dexterous Robotic Manipulation

Amazon Inc.; Zürich; Hybrid; CHF 140.000 - 210.000; Vollzeit

Zusammenfassung

Amazon seeks a Senior AI Engineer to advance reinforcement learning and deep learning for autonomous robots performing real-world doorstep deliveries. The role emphasizes learning from simulation and real data to refine dexterous manipulation, grasp synthesis, and end-effector control.

The candidate will address dynamic hand-offs, object interaction, and complex manipulation tasks while collaborating with the foundation model team and deploying models on hardware.

Qualifikationen

• Master's degree or higher in Engineering/Robotics or ML.
• 5+ years of industry or research experience; PhD applicable.
• Strong fundamentals in deep learning, RL, supervised and self-supervised learning.
• Robotics background with autonomy/manipulation experience.
• Production-level Python and system-level C++ coding.
• Experience deploying neural networks on hardware platforms.
• Experience with sim-to-real transfer and domain adaptation.

Aufgaben

• Develop RL algorithms for dexterous manipulation and robust end-effector control using multi-sensory inputs.
• Address real-world manipulation challenges such as varied packages and door-handling scenarios.
• Collaborate with the foundation model team to leverage simulated and real-world data.

Kenntnisse

Reinforcement Learning

Deep Learning

Python

C++

Robotics

MDPs

Sim-to-Real

Ausbildung

Master's degree or higher

PhD in Robotics/ML

Tools

TensorFlow

PyTorch

ROS

CUDA

Jobbeschreibung

Amazon seeks a Senior AI Engineer to advance reinforcement learning and deep learning for autonomous robots performing real-world doorstep deliveries. The role emphasizes learning from simulation and real data to refine dexterous manipulation, grasp synthesis, and end-effector control.

The candidate will address dynamic hand-offs, object interaction, and complex manipulation tasks while collaborating with the foundation model team and deploying models on hardware.