diff --git a/infra/website/docs/blog/feast-mlflow-kubeflow.md b/infra/website/docs/blog/feast-mlflow-kubeflow.md index 9b2111e2d44..3e15cbda26a 100644 --- a/infra/website/docs/blog/feast-mlflow-kubeflow.md +++ b/infra/website/docs/blog/feast-mlflow-kubeflow.md @@ -1,7 +1,7 @@ --- title: "Feast + MLflow + Kubeflow: A Unified AI/ML Lifecycle" description: Learn how to use Feast, MLflow, and Kubeflow to power your AI/ML Lifecycle -date: 2026-02-23 +date: 2026-03-09 authors: ["Francisco Javier Arceo", "Nikhil Kathole"] --- @@ -366,6 +366,11 @@ This is exactly the kind of insight MLflow's comparison interface is built for.

MLflow's comparison view showing three runs side by side with different feature subsets. The "Show diff only" toggle highlights how the features parameter varies across runs, making it easy to identify which combination of Feast features produces the best results.

+
+ MLflow metric charts showing accuracy, AUC, F1, precision, and recall grouped by num_features across three feature subsets +

MLflow's metric charts view visualizing accuracy, AUC, F1, precision, and recall across all feature selection runs, grouped by num_features. This chart makes it easy to spot how model performance changes as more Feast features are included.

+
+ Once you have identified the winning subset, the Feast registry ensures that only those features need to be materialized into the online store for production serving. ### Hyperparameter sweeps diff --git a/infra/website/public/images/blog/mlflow-feast-feature-selection-metrics.png b/infra/website/public/images/blog/mlflow-feast-feature-selection-metrics.png new file mode 100644 index 00000000000..b6ad16c3346 Binary files /dev/null and b/infra/website/public/images/blog/mlflow-feast-feature-selection-metrics.png differ