Description
If you believe the future of AI is just bigger GPUs and better backprop, this is not for you.
If you can imagine an embodied compute fabric made of many local units that carry their own state, exchange with nearby units, respond to bodily perturbation, retain traces, and change their internal structure when those changes improve or worsen viability, you probably should read this.
We are looking for a rare kind of compute architect: a systems philosopher who can code and design silicon for a new class of robots.
Not necessarily a chip architect.
Not someone to accelerate a neural network.
Not someone to put another model on an edge-AI board.
We are working on a different substrate problem.
Creature Algorithm is a body-coupled, morphogenic AI architecture. The system does not start from a trained model, reward function, central planner, or fixed network. It starts from local units, local consequence, viability, trace, structural change, and bodily interaction.
For robot prototypes, we can emulate this on traditional compute.
For the product path, we need to define and eventually build something more native to the architecture: a compute substrate where local state, local memory, local routing, local adaptation, and body-coupled signals are not afterthoughts, but the foundation.
We are looking for someone who can help us move from software-emulated local learning toward a dedicated body-coupled compute substrate. Carefully, practically, without jumping to custom silicon too early, and without collapsing the architecture back into conventional AI.
The first work is not to build the chip.
The first work is to help turn the already-defined substrate principles into engineering requirements, test architecture, and staged implementation.
what must remain local
what must become hardware-native
what can be emulated for now
what must never become centralized
what data the MVMP must generate
what the path is from PC/GPU emulation to FPGA, dense standby, sparse standby, and later custom silicon
If this sounds like the wrong job description for an AI chip role, that is probably the point.
We are not trying to build a faster brain for a robot.
We are trying to build the conditions under which a body can grow its own intelligence.
Requirements
The right person likely has serious experience in one or more of these areas:
* computational neuroscience
* developmental neurobiology
* embodied cognition
* neuromorphic systems
* event-driven compute
* FPGA / reconfigurable architectures
* ASIC architecture
* sparse local computation
* adaptive hardware
* embodied robotics compute
But the most important requirement is conceptual.
You need to understand why this is not a GPU/NPU problem.
Traditional chips compute.
Neuromorphic chips signal.
Morphogenic substrates build structure.
Very few engineers get the chance to work that close to first principles. If you are the right person, claim this rare privilege.
