Your mission & challenges
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.
What we can look forward to:
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)
