Multimodal Robotics Data Pipeline Engineer

Persona AI Inc; Houston (TX); On-site; USD 120,000 - 180,000; Full time

Job summary

Persona AI Inc. is hiring a Robotics Data Pipeline Engineer to architect scalable multimodal data pipelines and extract, augment, and align human dexterous manipulation data from large sensor-rich datasets.

You will build advanced post-processing algorithms for force analysis, infer hidden states from visual cues, and drive data curation for foundation models in demanding industrial environments.

Qualifications

• MS/PhD in Computer Science, Data Engineering, Machine Learning, Robotics, or related field.
• Strong Python expertise and PyTorch experience focusing on multimodal datasets.
• Experience with force/torque sensors or time-series data alignment with video.
• Proven video processing skills using OpenCV, FFmpeg, Decord.
• Solid 3D geometry knowledge (URDF, frames, transforms) for robotics data checks.
• Experience implementing data augmentation for CV and time-series data.

Responsibilities

• Architect end-to-end multimodal data pipelines from raw recordings to training-ready datasets.
• Design cross-modal validation systems to ensure video/state consistency across sensors.
• Coordinate hand-tracking, 3D reconstruction, and pose-tracking modules for robot trajectories.
• Implement robust data augmentation strategies to broaden expert trajectories.
• Develop teleoperation synchronization and unified state–action representations.
• Build tools for researchers to query, visualize, and audit datasets.

Skills

Python

PyTorch

Dataloader customization

Time-series data

OpenCV

Education

MS/PhD in CS/ML/Robotics

Tools

Open X-Embodiment

DROID

AgiBot World

Ray

Omniverse

Job description

Persona AI Inc. is hiring a Robotics Data Pipeline Engineer to architect scalable multimodal data pipelines and extract, augment, and align human dexterous manipulation data from large sensor-rich datasets.

You will build advanced post-processing algorithms for force analysis, infer hidden states from visual cues, and drive data curation for foundation models in demanding industrial environments.