Efference

Robotics teams collecting training data and running teleoperation or inference today rely on sensors and compute that are not built to work together, creating latency, jitter, and fragmented stacks across data collection, teleoperation, and inference. Scaling robotic perception requires hardware that is manufacturable at volume, not just lab prototypes. The page frames this as an unsolved problem in robotic-grade perception and compute at scale, implying current solutions fall short on fidelity, synchronization, and manufacturing readiness.

Efference is developing two devices: the M1, described as a robotics-grade device built for distributed data collection at scale, and the H1, a stereo system built for real-time cloud inference with foundation models. Its approach combines robotic-grade sensors that are hardware synchronized for low latency and zero jitter, integrated compute that aligns data collection, teleoperation, and inference under one stack, and hardware design optimized for manufacturing at scale. The M1 is slated for mass production in July 2026 and the H1 for limited production in September 2026.