Member of Technical Staff, Data Engineering

The role

This person will build and run data pipelines for large-scale multimodal datasets including video, robotics, and audio. The work involves tuning processing platforms and working with researchers to extract signals and manage the training mix.

What you will do

• Build and run data pipelines for large-scale multimodal datasets
• Pull useful signals out of raw data
• Work on the training mix by deduping and rebalancing clusters
• Work directly with researchers to turn model requirements into data strategy
• Improve the platform including storage, database layers, throughput, and deployment

Required qualifications

• Worked with real video and/or audio data
• Know what it takes to process video and/or audio at scale
• Roughly 1 to 5 years of relevant experience

Nice to have

• Computer vision experience, including classic CV like optical flow and feature or point tracking
• Experience with large-scale video or multimodal pre-training, data mixes, and curating data at that scale
• A background that mixes statistics or data science with engineering, plus hands-on video and image analysis
• Exposure to robotics data, human-movement data, or other emerging modalities like wearables or audio-heavy domains
• Audio experience specifically

About the company

Odyssey is an AI lab pioneering general world models through causal, multimodal systems. The team includes researchers from DeepMind, Tesla, Waymo, Meta, Apple, and Wayve.