Staff Robotics ML/Data Engineer: Scale Multimodal Data

Socket.dev; Town of Texas (WI); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Competitive compensation

Performance-based bonus

99% employer-covered medical benefits

Equity

Generous PTO and winter break

Job summary

Persona AI is hiring a Staff, Robotics ML/Data Engineer to architect and scale data pipelines turning multimodal recordings into training-ready datasets. You will extract and align force/torque signals with video, retarget human demonstrations to robot trajectories, and validate data with cross-modal checks.

This role drives how quickly our models learn dexterity in industrial robots. The position emphasizes Python/PyTorch, advanced video processing, 3D geometry, and end‑to‑end data pipelines in

Qualifications

• MS or PhD in Computer Science, Data Engineering, ML, Robotics or related field.
• Proficiency in Python and PyTorch with multimodal data loaders.
• Experience with force/time-series data from sensors.
• Video processing pipelines using OpenCV/FFmpeg/Decord.
• Strong 3D geometry, transforms, URDF knowledge for robotics data.
• Demonstrated data augmentation for vision and time-series data.

Responsibilities

• Force analysis and hidden state inference across video, proprioception, and signals.
• Kinematic retargeting and alignment for robot trajectories and validation.
• Advanced data augmentation to expand expert trajectories and robustness.
• Teleoperation synchronization and unified state–action representations.
• Build tooling to query, visualize, audit datasets and drive re-collection.
• Design end‑to‑end multimodal data pipelines for training-ready datasets.

Skills

Python

PyTorch

Time-series processing

Video processing

3D geometry

Data augmentation

OpenCV

Decord

Education

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

Tools

PyTorch

OpenCV

FFmpeg

Decord

Job description

Persona AI is hiring a Staff, Robotics ML/Data Engineer to architect and scale data pipelines turning multimodal recordings into training-ready datasets. You will extract and align force/torque signals with video, retarget human demonstrations to robot trajectories, and validate data with cross-modal checks.

This role drives how quickly our models learn dexterity in industrial robots. The position emphasizes Python/PyTorch, advanced video processing, 3D geometry, and end‑to‑end data pipelines in