DP-800 validates that you can design, build, secure, optimize and deploy databases
across Microsoft's SQL platforms and bring AI into them — external models, embeddings,
vector and hybrid search, and retrieval-augmented generation written in T-SQL. It is a developer's exam: you
are expected to read T-SQL, a dab-config.json, an SDK-style .sqlproj and a GitHub
Actions workflow, and spot what is missing or wrong.
How to use these notes
Start with SQL platforms & tooling — most traps on this exam are about where a feature exists and whether it is still in preview. Work the three skill pages in outline order, open each deep dive alongside the skill it supports, and keep the Configuration reference to hand while you practise. Drill Exam tips & caveats in the final week. These notes are a companion: pair every page with hands-on time in a free Azure SQL Database or a SQL database in a Fabric trial, and with the official Learn modules listed below.
Names you will meet in two versions
Azure AI Foundry is now Microsoft Foundry — the study guide uses the new name, while many SQL reference pages still say "Azure OpenAI in Azure AI Foundry Models". The Fabric-hosted engine is SQL database in Microsoft Fabric (not to be confused with the Fabric Warehouse or a lakehouse's SQL analytics endpoint). SQL Server 2025 is version 17.x in every "Applies to" banner. The full list is on the Exam tips page.
Skills at a glance
Two areas carry the same 35–40% and together make up three quarters of the exam; the AI area is the smallest at 25–30%, but its vocabulary — embeddings, chunking, ANN versus ENN, RRF — also turns up inside the other two.
Bars are drawn to scale against the largest weighting in the outline, 40%.
Skills measured
Design and develop database solutions
Tables, columnstore and specialized tables, JSON columns, constraints, sequences and partitioning; views, functions, procedures and triggers; advanced T-SQL; and building SQL with GitHub Copilot, Copilot in Fabric and MCP.
ExploreSecure, optimize, and deploy database solutions
Encryption, masking, RLS, permissions, passwordless access and auditing; isolation, Query Store and deadlocks; CI/CD with SQL Database Projects; Data API builder, Azure Monitor and change handling.
ExploreImplement AI capabilities in database solutions
External models and embeddings, chunking and embedding maintenance, full-text, vector and hybrid search with RRF, and retrieval-augmented generation from inside the database.
ExploreStart here, go deeper, review
SQL platforms & tooling
The three platforms the exam spans — SQL Server 2025, Azure SQL and SQL database in Fabric — the tools you build with, and which feature exists where, at what status.
Start hereAdvanced T-SQL deep dive
CTEs, window functions, the JSON and regular-expression families, fuzzy matching, graph MATCH, correlated queries and error handling — the code the exam asks you to read.
Read the codeVectors, search & RAG deep dive
One end-to-end build in T-SQL: an external model, chunks, embeddings, a vector index, exact and approximate search, full-text, hybrid ranking with RRF and a RAG call to a chat model.
Read the codeConfiguration reference
Every file the exam expects you to read: dab-config.json and the DAB CLI, the SDK-style .sqlproj and publish profiles, the GitHub Actions deploy workflow, Copilot instruction files and MCP server configuration.
Read the codeExam tips & caveats
Key numbers, GA-versus-preview by platform, the renames and the Learn-versus-docs discrepancies, commonly confused pairs, a scenario-to-answer lookup table and a checklist for the night before.
Drill theseBefore you book
Exam reality check
Expect the usual role-based formats — multiple choice, drag-and-drop ordering, hotspot drop-downs, yes/no
statement sets and case studies. Most questions describe a requirement and ask for the right object, option
or configuration: which encryption, which isolation level, which change-handling
mechanism, which search. The study guide warns that commonly used preview features can
appear, and on DP-800 that matters more than usual — vector indexes and VECTOR_SEARCH
are GA in Azure SQL and Fabric but still preview in SQL Server 2025. Try the
exam sandbox and the free
practice assessment before you book;
your registration confirms the exact exam time.
The exam in one picture
Almost every DP-800 question sits somewhere on this diagram. Schema reaches the database through a SQL project and a pipeline; applications and agents reach it through Data API builder; and the engine itself calls a model endpoint to embed, search and generate. If you can say which box a feature lives in — and which platforms that box supports it on — you can usually eliminate half the answers.
flowchart TB
DEV["Developer<br/>SSMS · VS Code MSSQL · GitHub Copilot"] -->|"SQL project · pull request"| CI["GitHub Actions<br/>build .dacpac · deploy"]
APP["Apps and AI agents"] -->|"REST · GraphQL · MCP"| DAB["Data API builder"]
CI -->|"SqlPackage publish"| DB
DAB --> DB
subgraph PLAT["One of three platforms"]
direction TB
DB["Database objects<br/>tables · JSON · graph · programmability"] --> SEC["Security and performance<br/>RLS · masking · Query Store"]
SEC ~~~ AI["AI in T-SQL<br/>embeddings · VECTOR_SEARCH · hybrid"]
end
AI -->|"sp_invoke_external_rest_endpoint"| LLM["Model endpoint<br/>Microsoft Foundry · OpenAI · Ollama"] Suggested study order
flowchart TB F["Foundations<br/>platforms · tools · availability"] --> S1["Skill 1 · Design and develop<br/>35–40%"] S1 --> S2["Skill 2 · Secure, optimize, deploy<br/>35–40%"] S2 --> S3["Skill 3 · AI capabilities<br/>25–30%"] S3 --> X["Exam tips and caveats<br/>final review"] S1 -.->|"alongside"| T["Advanced T-SQL<br/>deep dive"] S2 -.->|"alongside"| C["Configuration reference<br/>files and flags"] S3 -.->|"alongside"| V["Vectors, search and RAG<br/>deep dive"]
The official learning path
Course DP-800T00-A is a three-day instructor-led course, and its self-paced content is three Microsoft Learn learning paths with eleven modules — one path per skill area. Each module ends with an exercise or a knowledge check; read each one alongside the page it maps to.
| Learn module | Learning path | Maps to |
|---|---|---|
| Design and implement database objects with SQL | Design and develop database solutions | Skill 1 |
| Implement programmability objects with SQL | Design and develop database solutions | Skill 1 |
| Write advanced T-SQL code | Design and develop database solutions | Skill 1 · Advanced T-SQL |
| Implement SQL solutions by using AI-assisted tools | Design and develop database solutions | Skill 1 · Config reference |
| Implement data security and compliance with SQL | Secure, optimize, and deploy database solutions | Skill 2 |
| Optimize database performance | Secure, optimize, and deploy database solutions | Skill 2 |
| Implement CI/CD by using SQL database projects | Secure, optimize, and deploy database solutions | Skill 2 · Config reference |
| Integrate SQL solutions with Azure services | Secure, optimize, and deploy database solutions | Skill 2 · Config reference |
| Design and implement models and embeddings with SQL | Implement AI capabilities in database solutions | Skill 3 · Vectors & RAG |
| Design and implement intelligent search with SQL | Implement AI capabilities in database solutions | Skill 3 · Vectors & RAG |
| Design and implement RAG with SQL | Implement AI capabilities in database solutions | Skill 3 · Vectors & RAG |
Two habits that pay off more than extra reading
One: for every feature, learn its platform line — where it is GA, where it is
preview and what switch turns it on (PREVIEW_FEATURES, external rest endpoint
enabled, an MI update policy). Two: build the RAG pipeline once, end to end, in a
free Azure SQL Database: create an external model, chunk and embed a table, add a vector index, run a hybrid
query, then call a chat model with sp_invoke_external_rest_endpoint. An hour of that teaches
Skill 3 better than any page.
Disclaimer
These notes are an independent study aid for learning purposes only, not affiliated with or endorsed by Microsoft. SQL Server, Azure SQL and Microsoft Fabric ship changes often, and features move from preview to general availability between exam revisions — always verify every detail against the official Microsoft documentation linked throughout before relying on it.