Warehouse

Gold Feature Publishing

Curate silver views, publish gold datasets, and connect them to ML and Forecast training jobs.

Curate silver views, publish gold datasets, and connect them to ML and Forecast training jobs.

Business problem

Data science and analytics teams often maintain separate copies of “the dataset we use for modeling.” Without a governed publish step, feature definitions drift and production models train on stale tables.

Heimdall solution

Lake gold publish promotes validated silver tables to model-ready artifacts. ML, Forecast, and Loop read from the same gold catalog entry — one definition, many intelligence products.

Business impact

Feature stores live inside your warehouse layer. Models retrain on the gold version your team approved, with clear promotion history.

How it works

1

Validate in silver

Use Lab or silver views to confirm row counts, null rates, and business logic before promotion.

2

Publish to gold

Promote the dataset to gold with a stable name and schema your modeling team can depend on.

3

Train intelligence products

Select the gold table when creating ML, Forecast, or Loop projects — no manual export.

See Heimdall in action

Share how your team prepares data for modeling

Key benefits

Governed features

Gold is the approved version for modeling

One catalog

Same table powers ML, Forecast, and Loop

Reduced drift

Stop emailing CSV snapshots between teams

Faster retraining

Point models at updated gold when data refreshes

Clear ownership

Silver curation vs gold publish roles are explicit

Production confidence

Deploy models trained on cataloged gold only

Heimdall products for this workflow

Publish gold features from Lake

Curate silver, publish gold, and connect to intelligence products.