Data Engine
SUMMARY
The Data Engine is the data infrastructure of the Telekinesis Agentic OS. It standardizes how robotics data is represented, generated, collected, and organized into datasets that can be used to train and continuously improve Skills.
- Synthetic data seeds the first Skills.
- Skills are trained on the data.
- Trained Skills deploy and execute on real robots.
- Every run is captured as new data — which feeds back into training.
What is Data Engine?
Physical AI depends on data throughout its lifecycle - from model development and evaluation to deployment and continuous improvement.
The Data Engine provides the infrastructure for representing, generating, collecting, organizing, and reusing robotics data across perception, robot learning, and Physical AI workflows.
Data can originate from simulation, real-world systems, or existing datasets. The Data Engine brings these sources into a common data layer so they can be used consistently across training, evaluation, deployment, and future improvement.
Together, these capabilities support the Physical AI data lifecycle:
Generate / Collect → Organize → Train → Deploy → Improve
When to Use the Data Engine?
Use the Data Engine when you need to:
- Generate synthetic training data.
- Collect data from robots and deployed Skills.
- Create or work with robotics and vision datasets.
- Reuse existing datasets for training or evaluation.
- Standardize data exchanged between Skills, Agents, models, and logging systems.
What Does the Data Engine Provide?
The Data Engine provides the core infrastructure for generating, organizing, collecting, and standardizing robotics data across the Telekinesis Agentic OS.
- Data Loggers capture observations, actions, detections, telemetry, and outcomes from deployed systems.
- Data Generation generates synthetic training data for robotics and perception workflows.
- Dataset Operations create, load, write, inspect, configure, and visualize supported datasets.
- Dataset Catalog provides ready-to-use datasets for training, evaluation, and experimentation.
- Datatypes define the shared data contracts used across Skills, Agents, loggers, and datasets.
Guides
Data Loggers
Capture observations, actions, detections, telemetry, and robot interactions into structured datasets and logs.
Explore →Data Generation
Generate synthetic training data with Illusion, Extreme Domain Randomization, configurable randomizers, and distributed workers.
Explore →Dataset Operations
Create, load, write, inspect, configure, and visualize robotics datasets through a consistent interface.
Explore →Dataset Catalog
Browse ready-to-use datasets for training, evaluation, experimentation, and Physical AI development.
Explore →Datatypes
Strongly typed contracts for images, point clouds, poses, transforms, detections, meshes, trajectories, and more.
Explore →