Case Study: Artera cuts GPU inference costs 60% with Union.ai

Artera cuts GPU inference costs 60% with Union.ai

Artera, a pioneer in AI-guided cancer care, faced the challenge of scaling its GPU-heavy cancer test inference pipelines without letting costs spiral out of control. As the company expanded, each new pipeline and growth phase increased inference volume, making it critical to maintain accuracy and meet SLAs while controlling the expensive GPU time required. The hidden cost of container image-pull time on thousands of nodes was a significant tax on every task.

By implementing the Union.ai platform, Artera leveraged its Remote Image Builder and Cold Boot Optimization to reduce multi-minute image pulls down to milliseconds. This solution drastically cut the GPU node time per task. The results were a 60% reduction in GPU costs per inference task and a near halving of end-to-end run times. Union.ai's durable and self-healing runtime also provided the reliability needed to confidently scale analyses across 8000 nodes, allowing Artera to grow without sacrificing speed or dependability.

Artera