Job Summary
Job Title: Physical AI Engineer (Robotics & Agentic Systems)
We are looking for an engineer to build the Physical AI solutions using Robotics and edge devices on Google stack. You will integrate Deepmind Robotics model with Gemini 3 foundation models to create autonomous agents that perceive, reason, and act in dynamic industrial environments. Your work will involve moving Physical AI from simulation (MuJoCo-Warp) to production-grade hardware at scale.
Key Responsibilities
Key Responsibilities
• Model Integration: Deploy Gemini Robotics-ER (Embodied Reasoning) models to enable robots to understand natural language instructions and physical causality.
• Workflow Orchestration: Utilize the Gemini Enterprise Agent Platform to build agentic workflows that connect cloud-based reasoning with edge-based execution.
• Simulation & Training: Use MuJoCo-Warp and G4 VMs (NVIDIA Blackwell) to run high-fidelity simulations, generating synthetic data to train robots on complex edge cases.
• Hardware Abstraction: Work within Intrinsic Flowstate to develop hardware-agnostic "Skills" (robotic behaviors) that can be deployed across various robotic arms and sensors (Universal Robots, Foxconn, etc.).
• System Optimization: Implement the JAX AI Stack (JAX, Flax, Optax) for efficient, accelerator-oriented computation on Google TPUs and NVIDIA GPUs.
Skill Requirements
Technical Requirements
Category Stack Components
Foundation Models Gemini 1.5/3.0, RT-2 (Robotics Transformer), Gemini Robotics-ER
Software Platform Intrinsic Flowstate, ROS 2 (Robot Operating System), Python
Compute & Infra Google Cloud Vertex AI, Google Distributed Cloud (Edge), TPUs
Physics Engines MuJoCo-Warp, NVIDIA Isaac Sim integration
Data & ML JAX, XLA (Accelerated Linear Algebra), Orbax (Checkpointing)