The Indian Humanoid Startup Doing It on $3M | General Autonomy

In this talk

1. 2:16

   From consumer software to robotics

   Ahsan and Singh moved into robotics after building ShareChat and Moj, which reached a $5 billion valuation. Papers on diffusion policies and robot learning convinced them that end-to-end models could make a new kind of physical platform possible. They chose humanoids partly because Indian homes and workplaces often have stairs and other obstacles that wheeled robots struggle to cross.

2. 15:17

   Why build actuators in-house

   General Autonomy says actuator costs are central to making humanoids affordable, so it is building an actuator assembly line and pursuing robot dogs for inspection as an earlier market. Its second-generation actuator pairs an off-the-shelf frameless motor with custom precision gears and motor controllers, and is designed for up to 120 newton-meters of peak torque. The founders say custom hardware lets them match actuator performance to their models and address design constraints such as wiring and connectors.

3. 36:32

   Developing the robot and AI stack

   The company trains manipulation models from teleoperation data, using a diffusion transformer and online reinforcement learning to improve performance. The founders report initial task success rates of 70 to 80 percent, rising to 90 to 95 percent after online learning and corner-case work. They have demonstrated tasks including making coffee and carrying it to a conference room, but are currently prioritizing locomotion, navigation, and a more reliable next-generation humanoid.

4. 40:12

   Funding, ecosystem, and launch plans

   Ahsan says the team has raised about $3 million and has roughly 10 to 15 engineers. He argues India needs stronger robotics investment, shared testing and data infrastructure, and closer coordination between public programs and venture capital. General Autonomy expected to open developer preorders around October and put its first humanoids out in the first quarter of the following year; it expected robot dog deployments sooner, in the next quarter.

5. 46:28

   A billion-robot ambition

   The founders describe a long-term goal of putting a billion robots into the world within ten years, while acknowledging it is only a nonzero possibility. They expect humanoids to earn their place through useful work and value for money, and see specialized machines remaining better suited to many industrial tasks.