Innate is developing the AI brain for teachable home robots. From navigation to manipulation, interaction and planning, our robots already act entirely autonomously, and users can teach entirely new behaviors quickly. Our first robots are tailored for builders of Silicon Valley to quickly develop on top of and play with the state of the art in Embodied AGI
Former Stanford robotics graduate student. Former flight sciences intern at Archer. Also built high-powered rockets and underwater robots. Currently building the future of embodied AI with Innate
Building robots that feel alive. Stanford researcher in AI & Social Media and Human-Computer Interaction. MS in Applied Maths. Ex-founder of a media annotation platform. Ex-researcher at The Ocean Cleanup.
Innate builds AI brains for general-purpose robots, and affordable embodiments for consumers to experience the first AI robots.
Our first product, Maurice, is a mobile robot with a trainable arm, designed for Silicon Valley builders to play with, customize and build on. It runs our AI brain that can navigate, perform physical tasks, interact with people, and reason.
The brain is completely teachable: You can prompt it, chat with it, and teach it physical tasks.
It takes only 5 minutes for you to write another prompt for Maurice to behave very differently, for example, performing security at home:
In the past 3 months, we created our brain, an AI orchestrator leveraging several of the most recent research publications in AI for manipulation, navigation, reasoning and planning.
It combines 9 different models to create something that feels innately intelligent - like GPT, Gemini, Claude and Llama (Groq) got a body and learned to move. Maurice constantly understands its environment, makes decisions, remembers what is sees and where, and learns from you.
This system comprises:
🧠 A multi-agent orchestrator running specialized models locally and in cloud 🤖 Real-time manipulation models controlled by our own transformers
🦾 A full data pipeline to collect arm data, train in the cloud, run inference on the platform
💬 A 100% interactive platform, that can be steered vocally or through its prompts
📈 An eval pipeline for our brain’s decision making, to progressively make it better
Maurice learns your space, remembers context, and makes autonomous decisions about where, what to do next.
It’s also ready to build on:
All the compute required is onboard (uses a Jetson Orin Nano)
Shipped with a teleop arm to train it quickly
SDK to access the agent, steer it, give it new functions to call
And as a user, you can customize it, collect data quickly to train it, and even change the hardware.
This allows Maurice (and future robots) to learn several kinds of physical tasks.
Finally, Maurice is already reliable enough for us to do live, autonomous demos!
We're also developing a larger robot that can reach countertops. Our interoperable software ensures quick capability matching across all future embodiments using the Innate brain.
We're making teachable robots accessible to builders now, because we believe it is the right way to tailor robots people want, for a reasonable price. Furthermore, the data we collect along the way, teaching our robots, will be useful for all future robots.
Our first platform is for Silicon Valley builders who want to shape what teachable robots can do. Sign up for our waitlist on our website.
We are Vignesh and Axel, two former Stanford grad students and researchers in AI for robotics and AI for HCI. We met almost a year ago working on a previous idea, and through several iterations of this project, we have come to trust our shared abilities to face the hard problems ahead to make general-purpose robotics accessible to everyone.
This company embodies our vision of bringing a technology that will be a major game-changer for humanity, in the hands of everyone. So that robots are made for humans, taught by humans.
Us, 10mn before a successful live demo of our robot in front of 40 people, fixing an unexpected bug
The future of robotics is teachable. It’s time to build it!