All projects
Public demoAI Infrastructure / GPU

AI Model-to-GPU Flavor Calculator

Model-aware GPU VM sizing for LLM inference

Rackspace Technology

Built at Rackspace to support customer sizing and cost discussions.

A model-to-GPU VM sizing catalog that helps select GPU configurations from model size, precision, serving framework, context length and concurrency — to support configuration and cost discussions.

Problem & business context

Customers planning private LLM inference on rented GPUs needed a quick, explainable way to match a model and workload to the right GPU flavor, and to understand the cost implications.

My role & contribution

Designed and built the calculator.

What I built or designed

  • Sizing inputs: model size, precision, serving framework, context length and concurrency
  • Maps workloads to GPU VM flavors
  • Supports configuration and cost discussions with customers