Source description
About the role
Manifold Intelligence is building in physical AI: a data-efficient VLA stack for robust bimanual robot execution. We are currently operating in stealth and are opening exploratory conversations with exceptional technical builders. The current prototype includes a human-demonstration data pipeline, VLA training loop, dry-run evaluation tooling, and early robot hardware integration. I am a software founder and have built the first version of the stack end to end, including the demonstration-capture prototype, data collection workflow, training/deployment loop, and robot evaluation pipeline. Now I am looking to meet people who can complement me on robotics, hardware, ML, and real-world systems. This is an early, high-ownership role. Depending on fit and timing, it may become a full-time role, contract role, advisor relationship, or future founding-team position. You might be a fit if you have experience with one or more of: * Robotics or robot manipulation * Robot learning, imitation learning, VLA, or embodied AI * ROS2, robot control, calibration, or motion planning * Computer vision for robotics * Mechatronics, grippers, fixtures, tooling, or robot cells * Mechanical design / rapid prototyping * Data collection systems for physical-world tasks * Debugging messy real-world hardware/software systems What you would help with: * Move the stack from prototype hardware to reliable robot arms * Improve demonstration capture and data collection workflows * Build and debug robot execution/evaluation loops * Design grippers, fixtures, and robot-cell components * Help define safer action abstractions beyond raw motion deltas * Turn early physical-AI research into reliable bimanual robot demonstrations What we care about: * You build things, not just talk about them * You are comfortable with ambiguity * You can debug across software, hardware, and data * You have strong technical taste * You want high ownership at a very early stage * You are excited by general-purpose physical intelligence Manifold Intelligence is building in physical AI: a data-efficient VLA stack for robust bimanual robot execution. We are currently operating in stealth and are opening exploratory conversations with exceptional technical builders. The current prototype includes a human-demonstration data pipeline, VLA training loop, dry-run evaluation tooling, and early robot hardware integration. I am a software founder and have built the first version of the stack end to end, including the demonstration-capture prototype, data collection workflow, training/deployment loop, and robot evaluation pipeline. Now I am looking to meet people who can complement me on robotics, hardware, ML, and real-world systems. This is an early, high-ownership role. Depending on fit and timing, it may become a full-time role, contract role, advisor relationship, or future founding-team position. You might be a fit if you have experience with one or more of: * Robotics or robot manipulation * Robot learning, imitation learning, VLA, or embodied AI * ROS2, robot control, calibration, or motion planning * Computer vision for robotics * Mechatronics, grippers, fixtures, tooling, or robot cells * Mechanical design / rapid prototyping * Data collection systems for physical-world tasks * Debugging messy real-world hardware/software systems What you would help with: * Move the stack from prototype hardware to reliable robot arms * Improve demonstration capture and data collection workflows * Build and debug robot execution/evaluation loops * Design grippers, fixtures, and robot-cell components * Help define safer action abstractions beyond raw motion deltas * Turn early physical-AI research into reliable bimanual robot demonstrations What we care about: * You build things, not just talk about them * You are comfortable with ambiguity * You can debug across software, hardware, and data * You have strong technical taste * You want high ownership at a very early stage * You are excited by general-purpose physical intelligence