Build an ML model

Time to complete: ~10 min (samples) or ~20 min (gold, after ingest)

Before you start

Path	You need
Your data	Gold for ML, status ready
Sample	Signed-in account only

Open the wizard

1. Sidebar → Machine Learning (/ml).
2. Click New model.

Train from gold (your data)

1. Step Data source → under Published dataset (Gold), select your artifact.
2. Wait for schema load (target column pre-filled from gold metadata).
3. Step Target → confirm column to predict.
4. Step Train → name model → optional Advanced (search effort) → Train.
5. Training queue tab → open model when status is complete.

No upload step

Gold already passed validation (100+ rows, target, dtypes). The wizard loads from the artifact — you cannot upload CSV here.

Train from sample (explore only)

1. Step Data source → scroll below gold picker → Sample datasets.
2. Housing (regression, MEDV) or Iris (classification, Species).
3. Confirm target → name → Train.

→ Quick start: sample

After training

1. Open model at /ml/{id}.
2. Deploy → Generate API key.
3. Monitor on model page and Usage.

Troubleshooting

Symptom	Fix
Dataset missing from gold list	Data → Gold — confirm ML publish, status ready
Direct uploads are disabled	Use Lake → gold; only samples bypass Lake
Training failed	Check target column; republish gold if schema changed
Used zip ingest?	Landed in bronze — still need gold publish

Next steps

• Deployment
• API integration guide
• User journeys

Target column wrong?

Training fails or metrics look wrong. Fix gold metadata and republish if the target changed.