VinMotion

Key Responsibilities:

• Develop, train, and deploy reinforcement learning algorithms for robotic motion and manipulation tasks.
• Build and optimize simulation infrastructure to support large-scale policy training for general-purpose robots.
• Collaborate with the controls team to integrate learned policies into the robot’s existing control stack.
• Define performance metrics, test learned policies, and evaluate their effectiveness in real-world scenarios.

Requirements:

• Bachelor’s or Master’s, PhD’s degree in computer science, AI, Mechatronics, Control Engineering – Automation, or a related field
• Proficiency in writing production-quality code using PyTorch/Tensorflow/JAX.
• Familiarity with online and offline RL algorithms, such as PPO and SAC.
• Familiarity with simulation technologies (e.g., Gazebo, IsaacSim, NVIDIA Omniverse IsaacLab, Mujoco, Bullet)
• Experience in tuning hyperparameters, designing cost functions, and optimizing RL training processes.
• Knowledge of RL techniques, including domain randomization, curriculum learning, and reward shaping.
• Familiarity with machine learning evaluation tools like TensorBoard or Weights & Biases.

Preferred Qualifications:

• Experience in transferring policies learned in simulation to real robot hardware.
• Hands-on experience with massive parallelization training frameworks such as IsaacGym/ IsaacLab
• Hands-on experience in training locomotion policies for quadrupedal or bipedal robots. Proven experiences with RL applications in industry or real-world problems.