AI Engineer - Humanoid

Jobzhr; San Francisco, Northern (CA, KY); Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Equity

Founding team ownership

Hands-on role

Job summary

HUMANX, a San Francisco humanoid startup, seeks a Robot AI Engineer to join the founding team and ship autonomous capabilities for our humanoid. You will train, deploy, observe failures, collect data, and iterate rapidly on real-world tasks with a hands-on approach.

You will work with PyTorch, ROS/ROS2, and cutting-edge perception and policy models, driving sim-to-real pipelines and real robot deployment. Equity and ownership accompany this pivotal role.

Qualifications

• Strong experience in robot learning, manipulation, imitation learning, RL, or sim-to-real.
• Hands-on experience with PyTorch, transformers, multimodal/VLA models, or diffusion policies.
• Experience with RGB-D, 3D vision, point clouds, detection, segmentation, tracking, or pose estimation.
• Comfortable with Python/C++, Linux, ROS/ROS2, GPU computing, and real robot deployment.
• Ability to take models from training → simulation → physical robot → failure analysis → retraining.
• Ability to move quickly, experiment, and make good engineering tradeoffs with incomplete information.
• Strong ownership mentality and excitement about building difficult AI and robotics systems from scratch.

Responsibilities

• Develop learning-based policies for dexterous and bimanual manipulation.
• Build perception systems using RGB-D, 3D vision, segmentation, and pose estimation.
• Train policies using imitation learning, RL, diffusion policies, and human demonstrations.
• Adapt and deploy VLA and robotics foundation models on physical humanoids.
• Build simulation and synthetic-data pipelines with Isaac Sim, Isaac Lab, GR00T, Cosmos.
• Develop teleoperation and real-world data collection systems.
• Work on sim-to-real transfer, domain randomization, and policy generalization.
• Integrate learned policies with motion planning and whole-body control.
• Optimize neural policies for real-time deployment on robot hardware.
• Turn real-world failures into better training data and models.

Skills

Robot learning

Manipulation

Imitation learning

Reinforcement learning

Sim-to-real

Python/C++

Tools

PyTorch

ROS/ROS2

CUDA

TensorRT

Transformers

Job description

We’re an early-stage humanoid startup based in San Francisco and we’re looking for a Robot AI Engineer to join our founding team.

You’ll build the intelligence that makes our humanoid autonomous — enabling it to perceive the world, understand instructions, learn physical skills, manipulate objects, and reliably complete real-world tasks.

This is a hands‑on role. You’ll train models, deploy them on the robot, observe failures, collect data, retrain, and iterate quickly.

What you’ll do

• Develop learning‑based policies for dexterous and bimanual manipulation
• Build perception systems using RGB‑D, 3D vision, segmentation, pose estimation, and scene understanding
• Train policies using imitation learning, reinforcement learning, diffusion policies, and human demonstrations
• Adapt and deploy VLA and robotics foundation models on physical humanoids
• Build simulation and synthetic‑data pipelines with Isaac Sim, Isaac Lab, GR00T, and Cosmos
• Develop teleoperation and real‑world data collection systems
• Work on sim‑to‑real transfer, domain randomization, and policy generalization
• Integrate learned policies with motion planning and whole‑body control
• Optimize neural policies for real‑time deployment on robot hardware
• Turn real‑world failures into better training data and models

What we’re looking for

• Strong experience in robot learning, manipulation, imitation learning, reinforcement learning, or sim‑to‑real
• Hands‑on experience with PyTorch, transformers, multimodal/VLA models, or diffusion policies
• Experience with RGB‑D, 3D vision, point clouds, detection, segmentation, tracking, or pose estimation
• Comfortable with Python/C++, Linux, ROS/ROS2, GPU computing, and real robot deployment
• Ability to take models from training → simulation → physical robot → failure analysis → retraining
• Ability to move quickly, experiment, and make good engineering tradeoffs with incomplete information
• Strong ownership mentality and excitement about building difficult AI and robotics systems from scratch

Bonus

• Experience with humanoids, dexterous hands, or bimanual manipulation
• Experience with Isaac Sim, Isaac Lab, Isaac ROS, GR00T, Cosmos, CUDA, or TensorRT
• Experience with teleoperation and large‑scale robotics datasets
• Experience integrating learned policies with motion planning or whole‑body control

Why HUMANX

You’ll be joining the founding engineering team, with meaningful ownership over both the robot and the company (salary+equity).

“Pick up that bottle and pour the water into the cup,” you’ll build the intelligence that allows the robot to understand the task, perceive its environment, execute it, recognize failure, and improve.

You own the intelligence that makes it autonomous and useful.

Train it. Deploy it. Watch it fail. Collect the data. Retrain it. Watch it learn.

If you’ve always wanted to build a humanoid from the ground up, this is your chance.