Senior Reinforcement Learning Engineer,… at Figure AI

What you'd actually do

Develop, train, and deploy reinforcement learning algorithms for locomotion and manipulation tasks

Build simulation infrastructure to support the training of locomotion and manipulation policies for a general purpose humanoid robot at a large scale

Collaborate with the controls team to integrate policies into the existing control stack

Define, test, and evaluate performance metrics for learned policies

Skills

Required

PyTorch
online and offline reinforcement learning algorithms
PPO
SAC
hyperparameter tuning
cost function tuning
domain randomization
curriculum learning
reward shaping
TensorBoard
Weights&Biases

Nice to have

transferring policies learned in simulation to robot hardware
training locomotion policies for quadrupedal or bipedal robots

Figure is an AI Robotics company developing a general purpose humanoid. Our Humanoid is designed for corporate tasks targeting labor shortages and jobs that are undesirable or unsafe. We are based in San Jose, CA and require 5 days/week in-office collaboration. It’s time to build.

We are looking for a Senior or Staff level Reinforcement Learning Engineer. You will own the development, training, and deployment of new reinforcement learning algorithms for our humanoid robot as well as building infrastructure to support training policies at a large scale.

Responsibilities:

Develop, train, and deploy reinforcement learning algorithms for locomotion and manipulation tasks
Build simulation infrastructure to support the training of locomotion and manipulation policies for a general purpose humanoid robot at a large scale
Collaborate with the controls team to integrate policies into the existing control stack
Define, test, and evaluate performance metrics for learned policies

Requirements:

Confident writing production quality code in PyTorch
Familiar with online and offline reinforcement learning algorithms: PPO, SAC, etc.
Experience tuning hyperparameters and cost functions for these RL algorithms
Familiarity with common RL techniques such as: domain randomization, curriculum learning, reward shaping, etc.
Familiarity with general ML evaluation tools such as TensorBoard, Weights&Biases, etc.
Strong mix of industry and research experience, ideally 5-7+ years of experience

Bonus Qualifications:

Experience transferring policies learned in simulation to robot hardware
Experience training locomotion policies for quadrupedal or bipedal robots

The US base salary range for this full-time position is between $150,000 - $400,000 annually.

The pay offered for this position may vary based on several individual factors, including job-related knowledge, skills, and experience. The total compensation package may also include additional components/benefits depending on the specific role. This information will be shared if an employment offer is extended.