Chapter 4 of 6 · Databricks Generative AI Engineer Associate
Assembling and Deploying Applications
Pyfunc chains, Unity Catalog models, Vector Search, serving, ai_query and MCP.
- 4topics
- 20 minfast track
- 2diagrams and exhibits
- 22%of the exam
Topics in this chapter
Take them in order, or jump to the one you need.
- 1Chains, PyFunc Models and Unity Catalog RegistrationDatabricks GenAI Engineer: code a chain, wrap it in a pyfunc model with pre and post processing, pick RAG model elements and…Diagram5 min
- 2Vector Search Indexes and EndpointsDatabricks GenAI Engineer: Vector Search concepts, Delta Sync vs Direct Vector Access, standard vs storage optimized endpoints…Diagram5 min
- 3Serving, Access Control, Batch Inference and InterfacesDatabricks GenAI Engineer: serving on Foundation Model APIs, endpoint access to resources, batch inference with ai_query and chat…5 min
- 4MCP Servers, Memory, Prompt Versions and CI/CDDatabricks GenAI Engineer: managed, external and custom MCP servers, agent memory in Lakebase, Prompt Registry versions and…5 min
About this chapter
What this chapter covers
Assembling and Deploying Applications is 22% of the Databricks Generative AI Engineer Associate exam, the second largest chapter. It covers how the parts of a generative AI application are packaged, connected to data and tools, served to users and kept up to date.
The chapter has four parts: coding chains and pyfunc models and registering them in Unity Catalog; building and querying Mosaic AI Vector Search indexes; serving the application with Foundation Model APIs, access control, batch inference and a user interface; and the production pieces around an agent: MCP servers, memory, prompt versions and CI/CD.
Questions are mostly short scenarios in which the stated requirement (latency, cost, freshness, who may access what, quality signal) decides the answer. They are single-answer and some multiple-answer multiple choice, and short Python or SQL is shown as text. There are no exhibits.
How the topics fit together
You code the chain and wrap it as an MLflow model, then register it in Unity Catalog. Vector Search supplies its context, and Model Serving puts it behind an endpoint that users reach through an app or chat tool. A tested pipeline keeps prompts, indexes and versions moving safely to production.
- Chain questions ask how to code a chain or where pre and post processing belongs.
- Packaging questions ask which signature, input example, dependencies or resources a model needs.
- Vector Search questions ask which index type and endpoint type fit size, freshness, latency and cost.
- Serving questions ask about pay per token versus provisioned throughput, credentials and ai_query.
- Production questions ask about MCP server types, memory stores, prompt aliases and testing.
Look for the word that names the workload. "Every row of a table" points to ai_query, "updates from an application" points to Direct Vector Access, and "each user sees only their own data" points to on behalf of user authentication.
How to use these pages
Each topic page has the same shape, so you can skim for what you need.
- What it is: the idea in plain words.
- How it works: the mechanism, with tables and examples.
- What to remember for the exam: the facts to memorize and how questions are phrased.
- Common mistakes: the classic traps.
- Check yourself: three questions to test your understanding.
Ready for more practice?
The full Databricks Generative AI Engineer Associate question bank has 600 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Databricks. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.