Starry Recruitment Hiring! Full Time Physical AI Heterogeneous System Design Engineer in - Ricebowl

Physical AI Heterogeneous System Design Engineer

About the Role

We are seeking a Physical AI Heterogeneous System Design Engineer to develop next-generation robotic systems that combine AI, hardware acceleration, real-time control, and robotics. You will work across heterogeneous GPU + FPGA platforms, integrating multimodal AI models with robotic control systems to enable intelligent and low-latency autonomous manipulation.

Key Responsibilities

• Develop, train, and optimize Physical AI models using PyTorch, TensorFlow, and ONNX-based deployment workflows.
• Design and implement custom hardware accelerators using High-Level Synthesis (HLS), with a focus on low latency, resource efficiency, and timing closure.
• Develop closed-loop robotic manipulation and motion-control systems using ROS/ROS 2 and physics-based simulation environments.
• Integrate vision, force sensing, VLA model inference, robotic actuation, and system telemetry into end-to-end physical AI prototypes.
• Implement real-time sensor and motion-control interfaces on FPGA platforms, including PCIe and AXI4-based interconnects.
• Translate multimodal neural network architectures into C/C++ HLS implementations and ROS 2 middleware interfaces.
• Conduct simulation, hardware testing, performance optimization, and physical prototype demonstrations.
• Collaborate across AI, robotics, embedded systems, and hardware engineering teams to bring prototypes from concept to deployment.

Qualifications

• Master’s or Ph.D. in Electronic Engineering, Computer Engineering, Robotics, Computer Science, or a related field.
• Strong hands-on experience with PyTorch/TensorFlow, ONNX, and AI model optimization.
• Experience with ROS/ROS 2 and robotic motion-control systems.
• Strong understanding of multimodal AI architectures, including CNNs, ViTs, LLMs, VLMs, and VLA models.
• 3+ years of hands-on HLS experience, including hardware acceleration and FPGA-based system development.
• Demonstrated experience with timing closure, FPGA resource optimization, and hardware/software co-design.
• Experience with Physical AI, VLA models, robotic arms, and simulation/training environments such as MuJoCo, robosuite, or Isaac Sim.
• Strong programming skills in C, C++, and Python, including embedded drivers, hardware testbenches, and real-time applications.
• Experience integrating AI inference with sensors, actuators, and real-time robotic systems is highly preferred.

Ideal Candidate

The ideal candidate combines expertise across AI/ML, FPGA acceleration, robotics, and real-time systems and is comfortable moving from AI model development to hardware implementation and physical robot demonstration.