Skip to main content

Nostromo Labs

Nostromo Labs provides the data layer for physical AI, collecting, annotating, and packaging human demonstration data for embodied-AI teams. The company offers egocentric video collection, episode-level labeling with success/failure tags, and teleoperation-ready pipelines with kinematic chain data. Its managed teleop cell delivers the first batch within 48 hours from a nearshore facility.

Traction –/10
HQ unknown
310+ followers
  • Artificial Intelligence
  • Robotics
Updated 8 days ago

Funding

Funding rounds are not available yet.

Founders

1 founder

Jerry Weng

Exit Founder

Product

Problem

Embodied AI systems—humanoids, mobile manipulators, and industrial robots—cannot learn complex physical tasks from static text or simulated data alone. They require massive volumes of human demonstration data, yet most AI teams lack the infrastructure to collect, label, and package egocentric video and teleoperation data at the scale and quality needed for reliable model training.

Solution

Nostromo Labs provides the data layer for physical AI, collecting, annotating, and packaging human demonstration data into training-ready formats. The company offers egocentric video collection, episode-level labeling with success/failure tags, grasp events, object tracks, and language instructions, all cleaned and deduplicated. A managed teleoperation cell uses certified operators on client-matched hardware to generate demonstrations with in-session QC and scenario scripting. Every episode carries a documented QA trail, and data is delivered directly into the client's training pipeline.

Target Audience

Primary customers are foundation-model labs building vision-language-action (VLA) and world models, as well as warehouse, logistics, and industrial automation teams deploying mobile manipulation and factory robotics.

Features

  • Egocentric RGB video and multi-view video collection with speech and audio event capture in English and Spanish
  • Episode-level annotation including success/failure/recovery labels, temporal segmentation, grasp events, and keypoints
  • Teleoperation-ready pipelines with kinematic chain data such as joint angles and grip force readings per episode
  • Managed teleop cell with certified operators on client-matched hardware, including scenario scripting and in-session quality control
  • Per-episode QA trails ensuring each data point is verified before delivery
  • First batch delivery in 48 hours from a nearshore facility
This profile is AI-generated and may contain inaccuracies.