General-purpose bimanipulation, built on real-world data.
We build the data infrastructure and technology that enable general-purpose bimanipulation.
Our Mission
Closing the loop between data and deployment.
We’re a team of engineers and scientists building the data infrastructure for general-purpose, bimanual robotics. Data is only validated by closing the loop—training models and testing them in the physical world.
We work closely with our customers to turn that data into real-world deployment.
What We Build
The infrastructure behind Physical AI.
Platform
Loop
One platform that turns real-world demonstrations into deployable, task-specific robot policies — six composable modules, from collect to evaluate, powered by two GPU engines.
6
Modules
2
Engines
Foundation Model
CFG-1
Our first-generation robot foundation model for bimanual manipulation — trained on in-house human and robot data, adaptable to your task in hours.
10K+
Training hours
<50ms
Inference
Our Team
Built by people who’ve done this before.
Deep technical expertise in AI and robotics, shaped by years at leading institutions.
Latest
From the Blog
HABIT, the Dataset for Interactive Action Models
We release HABIT, the data for robots to coexist with humans: acting together, working side by side, and following human direction — and show the interactive behavior CFG-1 learns from it.
Teaching Robots to Remember
VLA with long-horizon memory: playing piano, shell games, and object rearrangement.
Teaching Robots to Coexist with Humans
At Config, we believe the hardest part of deploying robots isn't executing the task — it's operating in environments where people are present.