Teams at HowURun

Founding ML Performance Engineer will write custom CUDA kernels, optimize end-to-end model inference for sub-50ms latency, drive 10x stack performance (memory bandwidth, kernel fusion, scheduling), implement zero-copy multi-GPU/multi-node memory, own GPU utilization and memory management, profile and benchmark pipelines, and build performance tooling and standards.

Open call for exceptional candidates across functions. Seeking technical leaders and engineers who have built large-scale, low-latency, GPU-heavy inference or serving infrastructure, or top performers in GTM, design, devrel, or ops. Must own ambiguous problems end-to-end, set direction with limited scaffolding, and maintain high standards. Early-stage equity, competitive salary, full benefits, and top-tier AI tooling provided.

Founding ML infrastructure engineer to design and scale a GPU compute platform (1,000+ GPU clusters), build multi-tenant scheduling and marketplace infrastructure, own production reliability, architect feature stores/model registries and experiment tracking, and develop resource orchestration and scheduling to optimize throughput, cost, and latency.