Databricks agents
Agent authoring, tools, deployment choices, and lifecycle concepts.
Read official documentationOfficial platform documentation, terminology notes, and a concise glossary for the course.
Agent authoring, tools, deployment choices, and lifecycle concepts.
Read official documentationCurrent GenAI tracing/evaluation workflow; distinguish managed and preview features.
Read official documentationIndex types, setup requirements, synchronization, and current naming.
Read official documentationVersioned project/resource definitions and the validation/deployment lifecycle.
Read official documentationOfficial pipeline deployment walkthrough.
Read official documentationCurrent AI governance overview; verify controls for each invocation path.
Read official documentation| Older or posting term | Current teaching approach |
|---|---|
| Vector Search | Official documentation uses AI Search; verify SDK/import names separately. |
| Asset Bundles | Official documentation uses Declarative Automation Bundles. |
| AI Gateway | Current governance overview uses Unity Gateway; verify interface-specific controls. |
| Agent Evaluation | Use current MLflow 3 workflows and distinguish local from managed features. |
| DBRX | Understand the posting’s model example; select supported models based on task evidence. |
| Term | Meaning |
|---|---|
| FDE | Forward deployed engineer: an engineer working close to client workflows and delivery constraints. |
| RAG | Retrieval-augmented generation: supplying relevant evidence to a generation step. |
| Entitlement | A verified permission to access a resource or perform an action. |
| Idempotency | A repeated logical operation has the intended one-operation effect under a defined contract. |
| Trace | A connected record of the application operations for a request. |
| Held-out set | Evaluation cases reserved from the tuning/development loop. |
| p95 latency | A latency threshold covering roughly 95% of observed requests under a stated calculation. |
| Data grain | The meaning of one row in a table. |
| ADR | Architecture decision record: context, alternatives, choice, evidence, and consequences. |
| Runbook | An operational procedure for diagnosis, containment, recovery, and verification. |
Platform terminology references were checked on 5 October 2026. Product names, previews, regions, permissions, SDKs, and prices can change. Verify exact implementation details when performing a lab. The educational sequence and grading standards are independent design choices, not official certification requirements.
Managed Databricks labs have not been executed as part of this static website package. The bundled Python reference has local verification and clearly stated limits.