Software Engineer, Performance Tooling and Infrastructure

Who We Are

Nuro believes self-driving vehicles are the most immediate and profound opportunity for AI to drive positive change in the physical world. Safer streets, more time for what matters, and easier access to the world around us, that’s why we’re building a universal autonomy platform: self-driving for all roads and all rides. Founded in 2016, Nuro is a physical AI company developing Level 4 autonomous driving technology for a wide range of vehicles, use cases, and markets.

Powered by the Nuro Driver™, our universal autonomy platform enables the global mobility ecosystem to deploy autonomy at scale, from robotaxis and logistics fleets to personal vehicles. With years of real-world deployment experience and a flexible, partner-led business model, Nuro is working toward a future where millions of autonomous vehicles powered by our technology help make everyday life safer, easier, and more connected. Nuro has raised over $2B in capital from Uber, NVIDIA, Google, Softbank, Fidelity, T.

About The Role

Nuro leverages many different bench-top systems to evaluate and regression test different aspects of the software and hardware integration layer. This performance simulation platform includes systems At Nuro, every autonomy code change, from ML model updates to radius of map around the robot to number of evaluated trajectories, must be validated for real-time performance on actual robot compute hardware before it reaches the road. You will own the infrastructure that makes this possible.

Our Performance Simulation Platform is a hybrid benchmarking system: physical bench-top rigs running production robot compute (NVIDIA Thor platform), orchestrated by cloud-native infrastructure (Kubernetes, GCP), automated data pipelines feeding performance metrics into BigQuery and Grafana, pre/post simulation magic, custom tracing and profiling tools, and much much more. Engineers across the company rely on this platform daily to answer questions like: How will my new ML model affect contention on the GPU? How does a new data format impact onboard logging rate or network contention as more data might be flowing from through the system?

How much memory should be allocated for this new module, and how does it fit into the overall system budget? You'll be responsible for development, integration, and the evolution of this platform — from the bare-metal OS and networking layer through the job orchestration and CI/CD integration up to the data analysis and visualization layer. This is a high-ownership, high-autonomy role on a small team where your work directly gates the release velocity of the entire autonomy stack.

You'll be the technical DRI for the platform — setting the roadmap, making architectural calls, representing the platform's needs to the leadership team, and ensuring the system scales through multiple hardware generations. About the Work Benchmarking Infrastructure: Develop and maintain the job orchestration layer that schedules, executes, and validates autonomy performance benchmarks across a fleet of physical bench-top systems — integrated into CI/CD pipelines as merge-blocking and release-blocking quality gates. Platform Reliability Observability: Build monitoring, alerting, and self-healing automation for the bench fleet.

Proactively identify systemic risks — capacity bottlenecks, hardware degradation patterns, infrastructure single points of failure — before they become outages. Track utilization, failure rates, and capacity trends to ensure the platform scales ahead of organizational demand. Performance Data Pipelines: Design and build end-to-end data pipelines that capture fine-grained performance metrics (CPU/GPU utilization, memory bandwidth, E2E latency, scheduling jitter) from bench-top runs, process them at scale, and surface actionable insights through dashboards and automated regression detection.