LAYER 5 · DESIGN · STAGE 5

Cost optimisation for FDEs

Quick answer

Keeping model, cloud and engineering spend in line with value. A deployment that works but costs too much will not be renewed.

Practise it

  1. 1. Cache repeated model calls
  2. 2. Route easy requests to a smaller model
  3. 3. Report cost per resolved task

Where it sits

Stage 5 of the roadmap: Customer discovery. You can turn a vague brief into scope, a design and a plan a client signs.

Portfolio projects

Serve a small model on client infra