We make robots comfortable for living.
Proxemic is building the biological reward signal for physical AI. We teach autonomous systems the invisible shape of human space.
Request Early Access
RLHF aligned language.
RLHB aligns robots.
RLHB: Reinforcement Learning from Human Biology
Large language models learn what we want through thumbs-up and thumbs-down clicks (RLHF). But physical robots cannot wait for a rating. They operate in our personal space. We teach them to read the human autonomic nervous system directly from a camera.
Robots optimize what they can observe. Today, collision-avoidance frameworks plan kinematically flawless movements that remain blind to human psychological comfort.
By training student models on camera-synchronized autonomic ground truth (ECG, EMG, skin conductance), we produce a lightweight edge SDK. At deployment, the sensors are completely gone: a single onboard camera feed predicts human comfort in a few ms of compute.
Let's give robots common sense.
We partner with pioneering humanoid and collaborative robotics OEMs. Join us in building a future where robots are comfortable to be around.