Perception Deployment Engineer (CUDA)

Location: Hybrid in Foster City, CA; Boston, MA; Seattle, WA; or San Diego, CA

Company

Zoox develops autonomous vehicle systems and the software required to operate them safely and efficiently.

What you will do

• Design and develop production-grade, low-latency, memory-safe C++ and CUDA code for real-time perception algorithms on vehicle systems.
• Optimize large-scale multimodal sensor-fusion, LLM, and VLM models using quantization, pruning, and mixed-precision inference.
• Build model conversion and compilation pipelines with ONNX, TensorRT, and related tooling for edge deployment.
• Perform accuracy parity checks, recovery, and latency benchmarking between PyTorch models and compiled edge binaries.
• Develop custom ML operators and TensorRT plugins with efficient CUDA kernels to reduce latency and improve memory bandwidth on AI accelerators.

Requirements

• Production-level C++14/17/20 and Python programming skills.
• Experience developing concurrent, memory-safe, real-time inference code for edge devices.
• Deep expertise in model compression, including PTQ, QAT, pruning, and mixed-precision formats such as INT8, FP8, BF16, and FP16.
• Experience optimizing large-scale multimodal, LLM, VLM, or VLA models with efficient attention and KV-cache optimization.
• Extensive experience with model conversion and compilation pipelines, latency benchmarking, and model-quality parity evaluation.
• Proficiency in low-level AI accelerator programming, custom ML operators, TensorRT plugins, and CUDA kernel optimization.

Nice to have

• Experience with autonomous-driving perception algorithms such as temporal 3D object detection, BEV, and 3D Occupancy Networks.
• Experience processing vision, LiDAR, and radar data.
• Experience with end-to-end autonomous-driving paradigms, foundation models, and edge deployment technologies such as TensorRT-LLM.

Culture & Benefits

• Work as part of the Autonomy Software department and Perception team.
• Contribute to the development of next-generation autonomous system intelligence.
• Zoox is committed to building a team with diverse backgrounds, experiences, and skills.