Getting started: builders and publishers
Creating and publishing models, agents and prompts. What the role can do and a first-week path through the screens you will use.
For: Builder / Publisher · 1 min read
Your role on the platform
You build models, agents and prompts on the lakehouse and Vertex AI, and publish them so other business units can reuse them. Publishing puts your asset through the same checks every time: metadata, security scan, the responsible AI checklist, and validation by Model Risk Management where the risk tier needs it.
What you can do
- Build, train, package and register models, agents and prompts
- Version and publish assets into the Group catalogue
- Set reuse terms and chargeback rates
- Monitor own assets' adoption and value
Your first week
- 1Day 1: read the model pages of two flagship assets (Lead Priority Scorer, ETB Conversion Propensity) to see what a complete model card looks like.
- 2Day 2: open a notebook in the sandbox and request the datasets you need. Customer-level data needs an entitlement with a stated purpose.
- 3Day 3: read Policies, in particular model risk tiering and the fairness and explainability expectations for credit decisions.
- 4Day 4: draft a submission in Publish. Fill the responsible AI checklist honestly; reviewers read it.
- 5Day 5: check My assets for adoption and feedback, and the developer portal to see how consumers will call your asset.
Screens you will use most
Who to ask
The AI Centre of Excellence for publishing questions; Model Risk Management for validation scope.
Read next
Was this helpful?