Senior Staff/Principal Software Engineer - GPU System Software, Onboard Execution

Description

The company is building a robotaxi from the ground up and already runs it on public roads with riders. Every perception, prediction and planning decision runs on onboard compute that has fixed power and thermal limits, must be safe, and runs in real time.

GPU execution is among the company's most important engineering problems. Each new model generation needs more compute, while each new vehicle generation must cost less, draw less power and remain safe. Solving this is not only about tuning kernels. It requires a company-wide technical plan for how the full onboard AI workload is scheduled, shared, isolated and sized across GPUs, and the authority to carry that plan through ML, autonomy, middleware and hardware groups.

The role sits in the Software Performance organization, which covers end-to-end performance of the onboard software stack: profiling and observability, GPU compute and the compute platform. You will work closely with the Perception, Prediction, Planning, ML Infrastructure and Hardware teams and serve as a senior technical voice on onboard compute.

Responsibilities

• Design how perception, prediction and planning share the onboard GPUs, covering scheduling, multi-tenancy, isolation and predictable latency when workloads compete.
• Own a multi-year plan to lower onboard compute per vehicle by setting latency and utilization budgets and shipping the largest gains from models, runtimes, middleware and hardware.
• Guide choices for the next GPU and SoC platforms, including topology and memory budgets, backed by workload models and performance projections. Lead performance bring-up on each new platform.
• Set direction for the inference stack (TensorRT, CUDA, custom kernels, quantization) and work with ML teams so models are built for the hardware they run on.
• Lead and mentor senior engineers, get directors and VPs aligned through design documents, and decide when performance, safety and schedule conflict.

Requirements

• A BS in computer science, computer engineering or a related field, plus 10+ years in systems or performance engineering. Experience must include leading GPU or accelerator architecture work across teams or building and leading a performance engineering team.
• Deep knowledge of the CUDA platform, including programming model, runtime and driver, and of NVIDIA GPU microarchitecture from Ampere through Blackwell, including Orin and Thor.
• A track record of setting technical direction and delivering large, measurable end-to-end gains in latency, utilization or cost.
• Working knowledge of ML inference optimization (TensorRT, XLA, Triton, quantization, fusion) and expert-level C++ on Linux.
• Ability to get ML, systems, hardware and leadership teams behind one shared technical plan.
• Bonus: real-time or embedded GPU experience in autonomy, robotics, AR/VR, games or HPC.
• Bonus: production GPU multi-tenancy with MPS, MIG or priority streams.
• Bonus: new GPU architecture or platform bring-up, benchmarking methodology, or workload modeling and projections.

Conditions

Equal opportunity applies to all candidates.