What is MLOps?
MLOps is the discipline of deploying, versioning, monitoring and retraining machine learning models in production reliably.
Also called: machine learning operations
In more detail
It covers reproducible training, model and dataset versioning, evaluation gates in the deployment pipeline, drift monitoring once live, and rollback that does not require redeploying the application around it.
Why it matters
A model that performed well in a notebook will decay as the world it was trained on changes. Without MLOps that decay is discovered by customers rather than by monitoring.
How iLeaf approaches it
iLeaf versions models, prompts and retrieval indexes independently of application deploys, runs evaluations in CI so regressions fail the build, and plans drift monitoring and retraining into the first delivery.