Config Data Platform
Build, fine-tune, and evaluate robot models without the overhead of physical infrastructure—from raw data to final reinforcement, we handle the low-level layers so you can focus on the intelligence.
Build, fine-tune, and evaluate robot models without the overhead of physical infrastructure—from raw data to final reinforcement, we handle the low-level layers so you can focus on the intelligence.
Stage 1
Stage 1
Define Gripper
Define Gripper
Define Gripper
Manipulator and gripper
Manipulator and gripper
Stage 2
Stage 2
Design Human Gripper
Design Human Gripper
Design Human Gripper
Pre-made or custom
Pre-made or custom
Stage 3
Stage 3
Define Tasks & Environment
Define Tasks & Environment
Define Tasks & Environment
Tabletop or custom
Tabletop or custom
Stage 4
Stage 4
Design Data Strategy
Design Data Strategy
Design Data Strategy
Objects and motion
Objects and motion
Stage 5
Stage 5
Acquire Data
Acquire Data
Acquire Data
Human and teleop
Human and teleop
Stage 6
Stage 6
Validate & Label Data
Validate & Label Data
Validate & Label Data
Correctness and speed
Correctness and speed
Stage 7
Stage 7
Estimate Actions
Estimate Actions
Estimate Actions
On-prem or cloud
On-prem or cloud
Stage 8
Stage 8
Train Model
Train Model
Train Model
Multiple foundation model options
Multiple foundation
model options
Stage 9
Stage 9
Evaluate Model
Evaluate Model
Evaluate Model
Unit test and full test
Unit test and full test
Platform
Platform
User
User
CFG-1
Accelerate your development with CFG-1, our flagship foundation model. Built for native compatibility with our platform, it’s ready for immediate fine-tuning and deployment.
Accelerate your development with CFG-1, our flagship foundation model. Built for native compatibility with our platform, it’s ready for immediate fine-tuning and deployment.
Massive High Quality Data
Massive High Quality Data
100K Hours, 100% In-house, 100% human data
100K Hours, 100% In-house, 100% human data
Action estimation with micrometer-level precision.
Action estimation with micrometer-level precision.
2B Parameters
2B Parameters
Optimized for Efficient Training & Edge Inference
Optimized for Efficient Training & Edge Inference
<50 ms latency with RTX 5090
<50 ms latency with RTX 5090
Long Horizon Architecture
Long Horizon Architecture
Native 3-Minute Memory
Native 3-Minute Memory
Natively supporting long horizon tasks
while maintaining high resolution for short horizon.
Natively supporting long horizon tasks
while maintaining high resolution for short horizon.
Get a customized walkthrough from our experts.
Get a customized walkthrough from our experts.