Case Study: Wayve achieves >80% GPU utilization with NVIDIA Run:ai

Case Study: Autonomous Vehicle Company Ends GPU Scheduling “Horror”

Wayve, a London-based autonomous vehicle company developing embodied AI for self-driving cars, faced a GPU scheduling "horror": teams managed allocations in spreadsheets and in silos, leaving GPUs largely idle (about 25% average utilization despite 100% allocation). They engaged NVIDIA Run:ai and its Compute Management Platform to address these constraints.

NVIDIA Run:ai deployed its Compute Management Platform to pool GPUs, provide real-time visibility, and enable dynamic scheduling and provisioning on Kubernetes, boosting job throughput and accelerating AI experimentation. The result was efficient cluster utilization of over 80%, more than a 4× gain in infrastructure utilization, and a marked increase in the number of jobs running.

Wayve

Siddharth Sharma

Sr. Research Engineer