CrowdWorks
About
About CrowdWorks
CrowdWorks structures physical-AI training data into tasks, scenarios, subtasks and individual action segments. Its platform accepts recordings from teleoperation and simulation, with synchronised time-series and multi-camera viewers that let reviewers inspect how observations align with robot actions. Action-sequence labelling tools support the preparation of datasets for vision-language-action models.
The service combines automated quality checks with human annotation, validation and correction. Customers can commission data collection around particular behaviours, then evaluate models against real tasks and scenario variation. Workforce matching supplies contributors and specialist reviewers for these programmes. This joins data design, recording, detailed review and model evaluation in a managed workflow for robotics and autonomous-system developers.
Supply
Supply chain relationships
- MOBILUS
Outbound - robotics and autonomy data infrastructure - CrowdWorks provides agricultural-autonomy training-data annotation for MOBILUS's autonomous robot programme
Source
Timeline
Events
- 2026-09-21 - Contract Award - CrowdWorks wins MOBILUS agricultural autonomy data contract
The dataset uses pixel-level semantic segmentation for cultivated land, vegetation, farm roads and dirt tracks to improve field-boundary recognition. Obstacle labels capture size, position and distance so the autonomous-driving system can determine emergency-stop and slow-down behaviour. CrowdWorks will establish farm-specific annotation guidelines and apply its 2D, 3D LiDAR and multimodal data-processing experience to the project.
Source - 2026-07-01 - Product Launch - CrowdWorks launches robot precision-manipulation dataset for Physical AI training
The dataset was collected from 145 global remote operators controlling robot arms. It covers 12 manipulation scenario types and includes quality metadata such as completion, speed, path efficiency and safety scores. CrowdWorks positions the data for fixed robot arms and cobots in manufacturing, logistics and semiconductor environments.
Source
Context
Comparable profiles
- LabyrinthAI
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Encord
stack adjacency - Physical AI Data Infrastructure
- Mecka AI
stack adjacency - Physical AI Data Infrastructure
- micro1
stack adjacency - Physical AI Data Infrastructure
- Yitai Embodied
stack adjacency - Physical AI Data Infrastructure
- FORMOVE GmbH
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Vision Lab
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Sureform
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Xspark AI
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Kognic
stack adjacency - Physical AI Data Infrastructure · Physical AI Training Data
- Lightwheel
stack adjacency - Physical AI Data Infrastructure
- microagi
stack adjacency - Physical AI Data Infrastructure
Shared workflows
- shotwell.ai
workflow adjacency - Multimodal Robotics Data Labeling
- TwelveLabs
workflow adjacency - Multimodal Robotics Data Labeling
- AstrAlive
workflow adjacency - Multimodal Robotics Data Labeling
Context tags
- Physical AI data infrastructure - Stack Layer - Official Physical AI page documents an end-to-end multimodal data platform and managed pipeline.
- Multimodal robotics data labeling - Workflow - Official platform covers teleoperation, simulation, multi-camera and action-sequence data labelling.
- Physical AI training data - Stack Layer - CrowdWorks collects, labels and validates multimodal training data for robots, autonomous vehicles and vision-language-action systems.
Commercial
- CrowdWorks wins MOBILUS agricultural autonomy data contract
2026-09-21 · Order / contract · CrowdWorks won a data-annotation contract from MOBILUS to build training data for autonomous agricultural robots operating in unstructured farm environments.
Source