Heima — The First Humanoid Built for Developers

About Heima

The dark horse of
humanoid robotics

Heima (黑马) — Mandarin for "dark horse" — combines cutting-edge algorithms developed in Hangzhou, China with meticulous U.S.-based manufacturing to deliver unparalleled performance, reliability, and affordability in humanoid robotics.

2,000+

Household interactions learned through context-aware adaptation

$30B

Projected household robotics market by 2030

Q1 '26

Official A.D.A.M. platform launch date

100%

On-device processing for privacy and security

黑马

Heima — pronounced "Hey Ma'"

"Dark horse" — the unexpected contender that changes everything

The Platform

A.D.A.M.

Autonomous Dynamic Artificial Man — a general-purpose humanoid that evolves from developer platform today to the intelligence of the home tomorrow.

On-Device Processing

All computation happens locally. No cloud dependency, no data leaves your space. Privacy by architecture, not by policy.

Open SDK

Build, test, and share embodied intelligence on real hardware in minutes — not months. Clean APIs. Full documentation. No vendor lock-in.

Context-Aware Adaptation

A.D.A.M. learns household interactions and anticipates needs through a continuously evolving algorithm that adapts to your environment.

Whole-Body Motion

Natural, safe movement driven by advanced reinforcement learning policies trained in simulation and transferred to physical hardware.

Built for the Home

From handling chores and preparing meals to ensuring essentials are stocked — A.D.A.M. delivers proactive, personalized household assistance.

US-Made Hardware

Algorithms developed in Hangzhou. Manufacturing and assembly in the United States. Global talent, local precision.

1.6m

Height

60kg

Weight

25+

Degrees of Freedom

360Nm

Peak Torque

For Developers

Build embodied AI
on real hardware

A.D.A.M. ships with a clean, open SDK designed for rapid iteration. Whether you're a solo developer, a startup, or a research team — go from idea to running on real hardware in minutes.

Standardized platform — no more one-off hardware builds. Focus on your algorithms, not your actuators.

Sim-to-real pipeline — train RL policies in simulation and deploy directly to A.D.A.M.'s physical hardware.

Full-stack control — from high-level AI intent down to low-level motor commands through a unified interface.

adam_quickstart.py

from heima import Adam, Motion

# Connect to A.D.A.M.
robot = Adam.connect()

# Load a pre-trained locomotion policy
policy = Motion.load("walk_stable_v2")

# Execute whole-body motion
robot.execute(policy, duration=10.0)

# Stream real-time sensor data
for frame in robot.stream_sensors():
    print(frame.imu, frame.joint_torques)

# Deploy your own RL policy
my_policy = Motion.from_onnx("./my_model.onnx")
robot.execute(my_policy)

Our Team

Engineering experts &
serial entrepreneurs

Our founding team brings deep experience across robotics hardware, AI algorithms, control systems, and scaling technology companies.

Greg Wood

Chief Executive Officer

Serial entrepreneur and former regulatory change lead at Bridgewater Associates. Structured-finance attorney at Linklaters. MBA from Cornell Johnson, MIT Sloan Advanced Management Program.

Steve Guangcai

Head of Robotics Hardware

PhD in Solid Mechanics from Jilin University. Expert in battery management, power electronics, embedded systems, and multi-axis motion control. Leads hardware integration and manufacturing.

Ting Zou

Head of Algorithm Engineering (US)

MS from Carnegie Mellon, BA from Columbia. Former Amazon and Microsoft engineer. Expert in reinforcement learning, LLMs, and system optimization for embodied intelligence.

Eason Huang

Control Systems Lead

UW-Madison MS Statistics, UC Irvine BS Mathematics & CS. Serial AI entrepreneur. Owns A.D.A.M.'s real-time control layer and OS abstraction between high-level AI and physical execution.

Jack Xu

Mechanical Design Engineer

Jilin University MS. Former Senior Engineer at SANY Heavy Industry. 10+ patents, extensive experience in robotic systems and precision manufacturing. Owns production-grade mechanical design.

Xiang Yao

Simulation & RL Policy Engineer

Fudan University BS, University of Hong Kong MS. Hands-on experience tuning locomotion for humanoid and quadruped robots. Owns the simulation-to-real pipeline for RL policy transfer.

The first humanoid
built for developers

The dark horse of
humanoid robotics