Build local AI for real tasks.
Aetherseed combines an open AI framework, purpose-built hardware, and domain-specific knowledge modules. Use a finished solution, or build Aetherseed into your own product.
Three parts. One ecosystem.
Knowledge modules
The domain and capability layer. Gives the AI a concrete field, purpose, and limits. Ready-made, or developed with the customer.
Hardware
A purpose-built carrier board designed for the Raspberry Pi Compute Module 5 and expandable AI capacity, for local inference and integration into new products.
Framework
The shared software and system foundation. Open source, so engineering teams and product companies can build on it.
The architecture is the foundation. The hardware runs it. The knowledge module gives it a job. The product is what the customer builds or adopts.
One platform. Different fields.
A general language model knows a little about everything. A knowledge module makes it something bounded: a concrete field, built on verified sources, with clear limits on what it can answer and do. Change the module, and the same device solves a different task.
Or your own
If you have a field AI ought to handle locally, we develop the module together with you. A chiropractic module is a typical example: a bounded field, its own sources, a clear task.
Wonder Teacher is the first module being validated in use. The others are in development, and the selection above is representative rather than complete.
Four ways to use Aetherseed.
All four begin with a conversation, not an order. We are building production capacity in step with demand, and would rather take on a few customers properly than many halfway. Get in touch and we will work out what is possible for you.
Build a module with us
Companion with a customer-specific knowledge module, developed together with you.
Get in touch →Build a new product
Use Aetherseed hardware and the open framework as the basis for your own AI product.
Get in touch →Use your own hardware
Run the Aetherseed Framework on your own devices, with your own local or hybrid models.
Get in touch →Hardware built for local AI.
We develop our own carrier board because no existing board met our requirements for form factor, audio, battery, cooling, and swappable modules. The board is designed for integration into other products, not only for Companion.
- Compute Module 5
- Designed for the Raspberry Pi CM5, so the compute module can be swapped or upgraded.
- AI acceleration
- Room for a dedicated AI module. The photographs show a Hailo-10H fitted.
- Modular build
- Battery, compute, and AI module can be replaced separately, extending service life.
- Ready for integration
- Display, camera, I2C, UART, GPIO, expansion, USB-C, and micro-HDMI broken out on the board.
An open framework for local AI.
The Aetherseed Framework is the shared foundation under everything we build, and it is open source. Engineering teams can read it, build on it, and run it on their own hardware without asking our permission.
Beneath the framework sit three layers: Mustardseed governs how the agent behaves, AetherRoot holds local memory, and AetherSpark controls tool use through earned trust rather than fixed permissions. We are happy to walk through the technical detail when it is relevant.
The platform in finished form.
Companion is the reference product: proof that architecture, hardware, and swappable knowledge modules work as one coherent product. It is the simplest way to see the ecosystem work end to end — but it is not the only way to use Aetherseed.
Do you have a task AI ought to handle locally?
Talk to us about a knowledge module, product integration, or building on the ecosystem. We answer when we have something real to say.
Get in touch →