Open Role
Embodied AI Research Scientist (Robot Manipulation)
Singapore
Lead large-scale manipulation data pipelines and foundation model training.
Responsibilities
- Lead the development of large-scale robotic manipulation data pipelines, including data collection, management, and annotation systems.
- Define requirements and collaborate closely with engineering teams to build scalable and reliable infrastructure for embodied AI training.
- Design and optimize large-scale distributed training systems for robot manipulation foundation models.
- Develop data-driven learning approaches to improve robot generalization across tasks, objects, and environments.
- Research and deploy state-of-the-art techniques such as Reinforcement Learning, World Models, Sim2Real transfer, and Vision-Language-Action (VLA) models.
- Drive continuous improvement through closed-loop iterations of data, training, evaluation, and deployment.
Qualifications
- A deep and genuine passion for Embodied AI is essential.
- Master's or Ph.D. in Computer Science, Robotics, Artificial Intelligence, Automation, or related fields.
- 5+ years of experience in machine learning, robotics, or embodied AI research.
- Strong understanding of modern robot manipulation models such as Diffusion Policy, OpenVLA, π0, or related approaches.
- Extensive experience with PyTorch, TensorFlow, JAX, or similar frameworks.
- Experience building and operating large-scale distributed training systems.
- Strong background in robotic datasets, data infrastructure, and model evaluation.
- Publications in top-tier venues (CoRL, RSS, IROS, NeurIPS, ICML, ICLR, etc.) are highly preferred.