Design and develop our learned robotic manipulation software stack and algorithms
Develop robotic manipulation capabilities including but not limited to (re)grasping, pick-and-place, and more dexterous behaviors to enable useful work in both structured and unstructured environments
Model robotic manipulation processes to enable analysis, simulation, planning, and controls
Reason about uncertainty due to measurements and physical interaction with the environment, and develop algorithms that adapt well to imperfect information
Assist with overall software architecture design, including designing interfaces between subsystems
Ship production quality, safety-critical software
Collaborate with a team of exceptional individuals laser focused on bringing useful bi-ped humanoid robots into the real world
What You’ll Bring
Production quality modern C++ or Python
Experience in deep imitation learning or reinforcement learning in realistic applications
Exposure to robotics learning through tactile and/or vision-based sensors.
Experience writing both production-level Python (including Numpy and Pytorch) and modern C++
Proven track record of training and deploying real world neural networks
Familiarity with 3D computer vision and/or graphics pipelines
Experience with Natural Language Processing
Experience with distributed deep learning systems
Prior work in Robotics, State estimation, Visual Odometry, SLAM, Structure from Motion, 3D Reconstruction