The best harnesses for robotics teams. Data augmentation, training, and evaluation tools.
San Francisco
Joined April 2026
- Tweets29
- Following110
- Followers284
- Likes103
We're changing how teams can evaluate their robots in simulation. At scale. So you don't need to babysit a robot or a simulation environment.
Most robotics teams are blind. They run training for multiple hours and spend time writing down whether a roll-out worked or not.
This is expensive. And requires an engineer babysitting a robot or a simulation environment.
Kite ML (@trainrobotsfast) is here to change that.
Kite ML retweeted
We spent the last few weeks working with Cornucopia Biosciences on automating lab work.
Lowering the cost of science happens if we reduce the amount of manual work. With @trainrobotsfast, we built a model and dataset and made it open source.
Kite ML retweeted
Today scientists spend 40% of their time doing manual tasks. Robotics can push the frontier of scientific discovery. Cornucopia Bio is automating labs and using @trainrobotsfast to make that happen.
We've open-sourced a dataset, models, and a custom gripper for lab work here 👇
Cooking. @stisov16 and team trained a robot that picks up a ball and guards the goal.
They used Kite ML to augment datasets for different lighting conditions while keeping the rest of the environment the same.
Congratulations on such an incredible project!
We've truly revamped the landing page to communicate what we do best: help you train robots faster.
we've completely re-designed kite ml's (@trainrobotsfast) landing page to truly communicate what we do: create the best harnesses for robotics models to help teams train robots faster.
Kite ML retweeted
A team that's never touched a robot trained an SO-ARM in <3 hours, running ACT, SmolVLA, pi0.5 in parallel using Kite ML @trainrobotsfast and won the first place at the @AntlerGlobal Physical AI hackathon.
Insane.
Kite ML retweeted
Very excited about this. See you this weekend 👾
we are co-hosting one the biggest hackathons in physical ai in the world this weekend with @fdotinc and @Ryan_Resolution.
sign up at luma.com/0xbcm3kj?tk=OAIkJm
👩🏽🍳🤖
we are co-hosting one the biggest hackathons in physical ai in the world this weekend with @fdotinc and @Ryan_Resolution.
sign up at luma.com/0xbcm3kj?tk=OAIkJm
Kite ML retweeted
@trainrobotsfast has a new homepage. inspired by a kite's colors, yet minimalist. it gives you a sense of what you'll find inside without overwhelming you.
always be shipping.
ego-centric data in kite ml. that's the tweet.
Now, you can see ego-centric data in @trainrobotsfast.
In Kite you have the ability to both import data to train your VLAs and also visualize ego-centric data in our simulation environment.
Try it out. You might catch a glimpse of a robot dancing :)
world models have truly evolved the way robots can interact with more diverse and closer-to-real environments. if you want to try out marble 1.1, you can use it in our ide
AI’s next frontier is Spatial Intelligence, a technology that will turn seeing into reasoning, perception into action, and imagination into creation. But what is it? Why does it matter? How do we build it? And how can we use it?
Today, I want to share with you my thoughts on building and using world models to unlock spatial intelligence in this essay below. 1/n