Performance Engineer (Inference, Training & GPU)

World Labs is seeking a Performance Engineer to optimize the training and serving of their large-scale world models, ensuring they run as efficiently as possible on hardware. This individual contributor role involves identifying and eliminating bottlenecks across the entire stack, from low-level kernel optimization to fleet-wide serving efficiency, in close collaboration with AI researchers.

• Optimize inference and serving for latency and throughput
• Write and tune GPU kernels (CUDA, Triton)
• Optimize training throughput and GPU utilization
• Build performance models and profiling workflows
• Ensure numerical correctness across hardware and software changes
• Partner with researchers to productionize models