
Synapath AI is a video-first data infrastructure company for embodied AI that converts monocular internet video into robot-training data.
Synapath AI (Shutu Technology, Shenzhen) was founded in 2024 and is headquartered in Nanshan, Shenzhen. Its core product SynaData covers data capture, structured processing and standardized delivery. With Tsinghua University it co-developed the RoboWheel data engine, accepted to CVPR 2026; its HORA multimodal embodied dataset, with more than 150,000 manipulation trajectories, is open-sourced. In July 2026 it was named a Shenzhen Seed Unicorn.
SynaData is a data generation and delivery platform for embodied AI, producing six data lines: locomotion capture, single and dual-arm manipulation, full humanoid manipulation, egocentric video, human-object interaction (HOI) data and synthetic augmentation. It runs a dual capture engine combining distributed real-world sites with internet video, and applies video lifting and cross-embodiment retargeting; the company reports trajectory reconstruction error of about 0.4 to 0.5 cm and a 98% simulation transfer success rate.
According to CNR, its data products serve more than 60% of the hundred-billion-yuan valued embodied AI companies and have been purchased in volume by major technology manufacturers, with deliveries validated at humanoid robot companies and universities. It has completed angel+++ and Pre-A funding rounds with investors including Cornerstone Capital, Nanshan SEI and Suzhou Trend Capital.
No official updates yet — the chain engine will fill this timeline from verified official sources.