Tìm việc làm IT

REQUIREMENTS

Education & Experience

• 5+ years of experience in robotics data systems, ML data pipelines, or large-scale data infrastructure, including team or technical leadership experience.

• Hands-on experience building data pipelines for training ML / robotics models.

• Deep understanding of robot sensor data, including:

  • Vision (RGB, Depth)

  • Proprioception

  • Force / Torque

  • IMU

• Proven ability to scale data collection from small experiments to large robot fleets.

Technical Experience

• Experience with:

  • Humanoid or legged robots

  • Teleoperation systems

  • Imitation learning and reinforcement learning (RL) datasets

• Familiarity with:

  • ROS / ROS2

  • Cloud-based data infrastructure

  • Dataset versioning systems

• Understanding of:

  • Vision–Language datasets

  • Policy learning requirements, including temporal alignment and causality

Personality & Attitude

• Strong interpersonal and leadership skills.

• Proactive, responsible, service-oriented, and eager to learn.

• Strong communication skills with creative problem-solving ability and high attention to detail.

JOB DESCRIPTION

Key Responsibilities

Data Strategy & Ownership

• Define and own the end-to-end data strategy for humanoid robot AI training.

• Identify and specify data requirements for:

  • Manipulation learning

  • Locomotion and whole-body control

  • Task-level autonomy

  • Vision–language grounding

• Translate model and research requirements into concrete, executable data collection plans.

Data Collection & Pipelines

• Lead the development of scalable data collection pipelines, including:

  • Teleoperation and human demonstration systems

  • Autonomous data collection on real robots

  • Simulation-generated data pipelines

• Own policies and tooling for:

  • Experiment orchestration

  • Multi-modal data logging and synchronization (vision, proprioception, force, audio, language)

  • Versioned dataset generation and lifecycle management

Labeling, Quality & Optimization

• Design efficient data labeling strategies, including:

  • Human annotation

  • Programmatic labeling

  • Self-supervised and weakly supervised methods

• Drive dataset balancing, filtering, and validation to maximize model performance and generalization.

Infrastructure & Collaboration

• Partner closely with AI infrastructure and training teams to ensure:

  • High-throughput access to training data

  • Experiment reproducibility

  • Continuous dataset improvement pipelines

Team Leadership

• Build, manage, and mentor cross-functional teams, including:

  • Robot operators and data collectors

  • Data engineers

  • Data operations and QA teams