Simulation → Dataset
From a robot policy to a reusable dataset
What we learned collecting SO-101 demonstrations in Isaac Sim and exporting them to LeRobot.
A platform for your robotics ideas
We're building a place to create environments, collect data, train your own models, and test them—from simulation to real robots.
Start today with simulation, robot access, and recording.
01 / The platform we’re building
We provide the workspace and the tools. You bring the idea. Our goal is to connect every step of a robotics experiment in one place.
Set up the task you want a robot to learn. We're building toward reusable environments with your objects, different robots, and repeatable resets—so you can try an idea and keep improving it.
Demonstrate a task and turn useful attempts into a dataset. Our aim is to connect recording, dataset management, and training workflows, so you can teach your own model with your own data.
Bring a compatible model or a controller you've written. We want you to run it on a task, compare results, and learn from failures—in simulation first, with a path to real robot experiments.
Save demonstrations, connect a compatible controller, and compare simulation runs. Hosted training is planned.
See what you can try todayChoose a workspace. Define a goal. Make it your own.
02 / Start today
Try an idea in simulation.
Explore real hardware when you’re ready.
Create a scene, control an SO-101 arm, and record a demonstration. Use the simulation to explore a task without owning a robot.
See how a physical SO-101 moves and interacts with its workspace. Join the queue to control it and record your own attempts when it’s online.
03 / Research in the open
Simulation → Dataset
What we learned collecting SO-101 demonstrations in Isaac Sim and exporting them to LeRobot.
Why we’re building VeraPulse
From a student’s first experiment to a researcher’s next model, our aim is to make robotics easier to explore. Shared environments, robots, and data tools give more people a place to develop their ideas.
Room for different approachesStart with demonstrations and imitation learning. We’re designing the foundation to support reinforcement learning, scripted controllers, and other ways to teach and control robots.
Want to build with us? Get in touchNo. The simulation runs in a modern browser with WebGL support. You can also visit our shared physical robot through the remote console.
Currently, recordings are stored in this browser on this device. Signing in does not upload or back them up. Download important recordings before clearing browser data.
You can collect demonstrations, save datasets, and connect a compatible controller running on your own computer or server. Test it on our simulation task and compare saved results. Direct weight uploads and hosted training are not available yet.
Open the robot console to check the connection and cameras. When the robot is online, join the queue for a turn. The shared robot has one operator at a time.
Bring your curiosity. Build from here.