Jingxiang “JX” Mo is co-founder and chief executive of Gradient Robotics, which develops autonomous robots for AI infrastructure. Previously, he helped lead K-Scale Labs’ development of affordable, open-source humanoid robots, combining modular hardware with software and machine-learning systems that developers could adapt themselves.
Raised in Vancouver, Mo studied at McGill University and Nanyang Technological University, working on robotic manipulation and autonomous-vehicle navigation. At Flojoy, he led development of a graphical interface for programming industrial robots, testing their behaviors in simulation and deploying them onto physical machines.
In 2024, he collaborated with Kelsey Pool and Denys Bezmenov on Zeroth-01, a 3D-printable humanoid with a starting bill of materials around $350. The project integrated open hardware, robot operating-system components and reinforcement-learning environments into an accessible platform for experimenting with locomotion and sim-to-real deployment.
At K-Scale, Mo became head of product and engineering and helped develop K-Bot and Z-Bot: a roughly $9,000 research humanoid and a smaller robot sharing its control architecture. Their designs emphasized replaceable limbs, interchangeable end effectors, upgradeable computing and VR teleoperation. His AI Engineer World’s Fair presentation outlined three defining technical priorities:
- Open the complete robotics stack: Make hardware designs, electronics, firmware, software and learned policies accessible for outside developers to inspect and modify.
- Unify simulation and physical deployment: Give digital twins and real robots the same Python and Rust control interfaces, allowing policies to move between environments by changing their connection endpoint.
- Layer control and intelligence: Combine reinforcement-learning locomotion and whole-body control with higher-level vision-language-action models, while distinguishing available teleoperation from still-unrealized household autonomy.
Mo also co-authored EdgeVLA, which uses smaller model backbones and non-autoregressive action prediction to accelerate robotic inference on constrained hardware. Its authors reported a sevenfold inference-speed improvement and released an open-source implementation.
After K-Scale, Mo founded Gradient through the South Park Commons founder community, applying his robotics experience to building and operating data-center infrastructure.
