Zoox is building a purpose-built robotaxi from the ground up and running it on public roads today, with riders. Every perception, prediction, and planning decision runs on an onboard compute: a fixed-power, fixed-thermal, safety-critical, real-time system.
That makes GPU execution one of the most consequential engineering problems in the company. Each new model generation wants more compute. Each new vehicle generation needs to cost less, use less power, and stay safe. Closing that gap is not a kernel-tuning exercise. It takes an end-to-end technical vision for how the whole onboard AI workload is scheduled, shared, isolated, and sized across GPUs, and the authority to drive that vision across ML, autonomy, middleware, and hardware organizations.
You will join Zoox's Software Performance organization, which owns end-to-end performance of the onboard software stack: profiling and observability, GPU compute, and compute platform. You will work closely with Perception, Prediction, Planning, ML Infrastructure, and Hardware teams, and you will be a senior technical voice for onboard compute across the company.
We are looking for the technical leader who sets that vision. You will define how Zoox runs AI on the vehicle, and you will own the strategy for doing more with less compute on every vehicle we build next.
In this role, you will:
- Architect onboard GPU execution. Define how perception, prediction, and planning workloads share onboard GPUs, including scheduling, multi-tenancy, isolation, and deterministic latency under contention.
- Drive compute reduction. Own the multi-year roadmap to cut onboard compute per vehicle by setting latency and utilization budgets and landing the biggest levers across models, runtimes, middleware, and hardware in real releases.
- Shape hardware and platform direction. Guide next-generation compute decisions (GPU and SoC selection, topology, memory budgets) with workload models and performance projections that justify them, then lead performance bring-up of each new platform.
- Lead inference performance at scale. Set direction for the inference stack (TensorRT, CUDA, custom kernels, quantization) and partner with ML teams so models are designed for the hardware they ship on.
- Provide technical leadership. Lead and mentor senior engineers, align directors and VPs through design docs, and make the hard calls when performance, safety, and schedule conflict.
Qualifications
- BS in computer science, computer engineering, or related field and 10+ years in systems or performance engineering, including leading GPU or accelerator architecture across teams, or building and leading a performance engineering team.
- Deep CUDA platform (programming model, runtime, and driver) and NVIDIA GPU microarchitecture expertise (Ampere through Blackwell, Orin, Thor).
- Proven record 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 C++ on Linux.
- Ability to align ML, systems, hardware, and leadership on a shared technical plan.
Bonus Qualifications
- Real-time or embedded GPU experience: autonomy, robotics, AR/VR, games, or HPC.
Production GPU multi-tenancy (MPS, MIG, priority streams).
New GPU architecture or platform bring-up, benchmarking methodology, or workload modeling and projections.