AI engineering glossary · MLOps, LLMOps cloud
What is champion-challenger?
Champion-challenger: A method where the current production model, the champion, runs against one or more newer candidates, the challengers, on live or replayed traffic. A challenger replaces the champion only if it proves better on agreed measures.
Champion-challenger: an example
A bank keeps its scoring model as champion while a new gradient-boosted challenger scores the same applications in the background for a month.
Related terms
More MLOps, LLMOps cloud terms
- CI pipeline
- CI/CD
- Cloud region
- ConfigMap
- Configuration drift
- Container layer
- Container runtime
- Continuous delivery
Where you learn it
Champion-challenger is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.