Markov Robotics

Markov Robotics is a US startup headquartered in San Francisco, founded in 2025, building general-purpose intelligence for robots. Its technology is based on a world model conditioned on high-fidelity modalities such as pressure and proprioception, enabling closed-loop control and limiting the accumulation of errors. The company aims to solve zero-shot general-purpose dexterity: a single policy intended to run across embodiments — from dexterous hands and grippers, through single arms, to humanoid robots. The model learns from human demonstrations and video, then is deployed through hardware partners, feeding an iterative improvement loop on real-world deployment data. The founding team (Atharva Gundawar, Greg Tarr, Diego Caples) met at AGI, Inc. The company raised a $5M seed round with participation from Pebblebed.