-
Multimodal Learning Architecture: You design and build systems that fuse camera, depth, tactile, IMU, language, and robot-state inputs into executable robot actions. You define how these modalities combine and translate into policies that run on our dexterous hands.
-
Simulation Infrastructure & Data Pipelines: You build and maintain GPU-accelerated simulation environments and the data pipelines that make policy training scalable.
-
Sim-to-Real Transfer: You lead the engineering work that closes the gap between simulation and physical hardware.
-
Team Leadership & Engineering Direction: You translate research problems into concrete milestones, guide engineers, and work across ML, robotics, and hardware teams to ship learning systems that run on physical robots.
-
6 or more years of experience in computer science or a related engineering field, with meaningful time delivering AI systems on physical robotic hardware
-
Hands-on experience leading or co-leading the design of multimodal manipulation systems combining vision, language, tactile, and proprioceptive inputs
-
Proven track record building simulation infrastructure (Isaac Lab and Isaac Sim, or MuJoCo) for reinforcement learning and sim-to-real transfer
-
Deep practical knowledge of imitation learning (including diffusion policies), deep RL, and hybrid learning approaches on real robot hardware
-
Experience with data pipelines for heterogeneous, high-frequency sensor data: teleoperation, tactile, vision, depth, and robot state
-
Strong Python and C++, with experience in ROS2 and embedded or real-time systems
-
Ability to translate ambiguous research problems into concrete engineering milestones and to grow junior engineers
Nice to have:
-
Experience with high-DOF, tendon-driven, or tactile-heavy dexterous hands
-
Familiarity with VLA or vision-language-action architectures and large-scale pre-training workflows
-
Contributions to the robotics or AI research community (ICRA, IROS, CoRL, NeurIPS)