Robotics AI Data Engineer for Multimodal Training

Persona AI Inc; Georgia; On-site; USD 120,000 - 170,000; Full time

Benefits offered by this job

Performance bonus

Medical benefits

Equity

Paid time off

Winter break

Job summary

Persona AI Inc is seeking an AI Engineer – Robotics Data Preprocessing to architect and scale ingestion pipelines turning multimodal data (egocentric video, IMU, haptics) into training-grade assets for humanoid robots. You will drive cross-modal validation, data augmentation, and telemetry-aligned data curation.

The role emphasizes building scalable data layers, retargeting demonstrations to robot trajectories, and advancing model dexterity through high-quality datasets.

Qualifications

• MS/PhD in CS/ML/Robotics or related field.
• Strong Python and PyTorch with multimodal data loaders.
• Experience with force-torque sensors and time-series data.
• Video processing pipelines with OpenCV/FFmpeg.
• Knowledge of 3D geometry, transforms, URDF.

Responsibilities

• Design cross-modal validation systems for video, proprioception, force/haptic signals, and language annotations.
• Retarget and align demonstrations to robot trajectories with kinematic feasibility checks.
• Implement robust data augmentation to expand expert trajectories for robust models.
• Coordinate teleoperation synchronization across embodiments and sensors.
• Build tooling to query, visualize, and audit datasets and drive curation rules.
• Develop end-to-end multimodal data pipelines from raw recordings to training-ready datasets.

Skills

Python

PyTorch

Time-series data processing

Video processing

3D geometry

Data augmentation

Robotics data

NVIDIA robotics stack knowledge

Education

MS or PhD in Computer Science, Data Engineering, ML, Robotics

Tools

OpenCV

FFmpeg

Decord

Open X-Embodiment / DROID / EgoDex (bonus tools)

Job description

Persona AI Inc is seeking an AI Engineer – Robotics Data Preprocessing to architect and scale ingestion pipelines turning multimodal data (egocentric video, IMU, haptics) into training-grade assets for humanoid robots. You will drive cross-modal validation, data augmentation, and telemetry-aligned data curation.

The role emphasizes building scalable data layers, retargeting demonstrations to robot trajectories, and advancing model dexterity through high-quality datasets.