Thiết kế và triển khai thuật toán thao tác cho robot humanoid tại VinMotion
Yêu cầu
Bằng cấp S3, S2 hoặc S1 trong CNTT, AI, Robotics, Mechatronics
Kỹ năng lập trình Python và các framework Deep Learning (PyTorch, TensorFlow, JAX)
Nắm vững các thuật toán RL online/offline (PPO, SAC)
Kinh nghiệm với nền tảng mô phỏng robot (NVIDIA Isaac Gym, Isaac Lab, MuJoCo)
Trách nhiệm
Thiết kế và triển khai thuật toán thao tác cho robot humanoid
Huấn luyện và tối ưu hóa mô hình Deep Learning và Reinforcement Learning
Nghiên cứu và triển khai các thuật toán RL online/offline để cải thiện độ chính xác
Tối ưu hóa đường ống huấn luyện RL và kỹ thuật kỹ thuật phần thưởng
Có thì tốt
Kinh nghiệm thực tế với tác vụ thao tác robot
Kinh nghiệm triển khai thuật toán trên hệ thống robot vật lý
Kinh nghiệm với tay robot có cảm biến xúc, servo thị giác, thao tác hai tay
Tóm tắt bằng AI từ JD gốc, đọc JD gốc trước khi ứng tuyển.
Location:
Gia Lam, Hanoi
About Us:
VinMotion is at the forefront of robotics innovation, developing advanced humanoid robots designed to push the boundaries of motion intelligence. We are looking for talented AI for Manipulation Engineers to develop, train, and deploy advanced AI and Reinforcement Learning (RL) algorithms for humanoid robot manipulation. This role will focus on dexterous manipulation, grasping, bimanual coordination, visual and tactile servoing, and complex object interaction to enhance the robots' ability to perform precise and adaptive manipulation tasks in real-world environments.
Key Responsibilities:
Design, implement, test, and deploy manipulation algorithms for humanoid robots with high degrees of freedom (DOF) and dexterous robotic hands.
Develop manipulation capabilities for complex tasks, including grasping, picking and placing, bimanual manipulation, visual and tactile servoing, and robotic assembly.
Train and optimize deep learning and Reinforcement Learning (RL) models to enable robust manipulation under uncertainty, sensor noise, and partial observations.
Research and implement advanced online and offline RL algorithms to improve manipulation accuracy, adaptability, and task execution.
Develop and optimize manipulation policies using simulation-based training and real-world robotic experiments.
Collaborate with cross-functional engineering teams to integrate manipulation algorithms into the full robot autonomy stack.
Define performance metrics, design evaluation frameworks, and validate learned policies through simulation and real-world testing.
Optimize RL training pipelines through hyperparameter tuning, reward engineering, and training performance improvements.
Research and apply state-of-the-art AI and RL techniques to advance dexterous manipulation and humanoid robot capabilities.
Requirements:
Bachelor's, Master's, or PhD degree in Computer Science, Artificial Intelligence, Robotics, Mechatronics, Control Engineering and Automation, or a related field.
Strong programming skills and experience developing production-quality code using Python and deep learning frameworks such as PyTorch, TensorFlow, or JAX.
Familiarity with online and offline Reinforcement Learning algorithms, particularly Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC).
Hands-on experience with robotic simulation platforms such as NVIDIA Isaac Gym, Isaac Lab, and MuJoCo.
Experience in hyperparameter tuning, reward engineering, and optimizing RL training pipelines.
Good understanding of traditional and learning-based robot manipulation approaches, including their advantages, limitations, and trade-offs.
Strong understanding of robot manipulation fundamentals, including grasping, dexterous manipulation, motion planning, and robotic control.
Good knowledge of recent developments, research, and frameworks in Reinforcement Learning for robotic manipulation.
Strong analytical thinking, problem-solving skills, and the ability to conduct independent research and development.
Preferred Qualifications:
Hands-on experience with robotic manipulation tasks, research, or real-world robotics projects.
Experience developing and deploying manipulation algorithms on physical robotic systems.
Experience working with dexterous robotic hands, tactile sensors, visual servoing, or bimanual manipulation.
Familiarity with Sim-to-Real transfer techniques for robotic manipulation.
Research publications in relevant fields, including Reinforcement Learning, Robot Manipulation, Robotics, or Machine Learning.
Contributions to open-source robotics or AI projects are an advantage.