About the role
Responsibilities:
- Architect the onboard dual-compute payload: high-performance compute (NVIDIA Jetson AGX Orin 64GB) for VLA/vision inference alongside real-time MCUs/x86 controllers for motor execution.
- Optimize model inference latencies using TensorRT, TensorRT-LLM, FP8/INT4 weight quantization, and zero-copy shared memory IPC.
- Architect hardware synchronization using IEEE 1588 PTP for 3D LiDAR, stereo cameras, and high-frequency IMUs.
- Engineer zero-copy inter-process communication pipelines between deep learning inference runtimes and ROS 2 middleware.
- Optimize onboard power distribution, dynamic thermal management (MAX-N operating modes), and system hardware watchdogs.
- Establish deterministic sensor synchronization across 3D LiDAR, RGB-D cameras, and IMUs using PTP (IEEE hardware timestamping).
Requirements:
- Bachelor's or Master's in Computer Engineering, Electrical Engineering, or Computer Science.
- 5+ years of embedded systems experience with NVIDIA Jetson platforms, CUDA programming, kernel optimization, and embedded Linux.
- Demonstrated track record of deploying quantized neural network pipelines on production robotic edge hardware.
- Jetson Linux (L4T), TensorRT-LLM, CUDA, C++, zero-copy shared memory, IEEE 1588 PTP, hardware watchdogs.