AI-enabled database training: when AI moves into SQL
For years, adding AI to an application meant copying data to another service: a vector database on one side, the relational database on the other, and scripts to keep them in sync. Every copy added risk, delay and maintenance work.
Microsoft SQL platforms have changed that. SQL Server, Azure SQL and SQL databases in Microsoft Fabric can now store vectors, call external models and combine semantic search with regular filters in a single query. The Develop AI-enabled database solutions (DP-800) course starts from there: the SQL developer becomes a key player in AI projects. For organizations in French-speaking Switzerland that keep sensitive data in SQL Server, it is also a way to bring AI to the data rather than the other way round.
Vector search, hybrid search and RAG inside the database
At the heart of this AI-enabled database training is intelligent search. You learn to choose between full-text search, semantic vector search and hybrid search depending on the question being asked. You also prepare the data: which columns belong in an embedding, how to split long text into chunks, and how to keep embeddings current when rows change.
Then come the design decisions that matter in production: the vector data type, vector indexes, exact (ENN) or approximate (ANN) search, and merging rankings with Reciprocal Rank Fusion. Retrieval-augmented generation (RAG) closes the loop: the database supplies the context, the language model writes the answer, and you stay in control of what gets sent to it.
Developing in SQL with GitHub Copilot and MCP servers
AI does not only serve end users, it also changes how code gets written. The course shows you how to enable GitHub Copilot and Copilot in Fabric, write instruction files that enforce your conventions, and connect the assistant to MCP servers that reach SQL Server or a Fabric lakehouse.
That productivity boost has a flip side: an assistant wired to a database can see more than it should. The curriculum therefore covers the security impact of these tools, so you keep control over the data they expose.
Securing, tuning and shipping like a software team
A database that feeds an AI application is still a critical database. The course covers encryption with Always Encrypted, dynamic data masking, row-level security, auditing and passwordless access to models through managed identities. On the performance side, you read execution plans, work with Query Store and resolve blocking and deadlocks.
Deployment follows software engineering practice: SQL Database Projects in source control, pull requests, schema drift detection and CI/CD pipelines. Data API builder then exposes your tables, views and stored procedures over REST and GraphQL, with no hand-written API layer.
Microsoft Certified: SQL AI Developer Associate certification
The course prepares you for the DP-800 exam (Developing AI-Enabled Database Solutions), which leads to the Microsoft Certified: SQL AI Developer Associate certification. The exam lasts 120 minutes, requires a score of 700 and covers three areas:
- designing and developing database solutions (35 to 40%)
- securing, optimizing and deploying those solutions (35 to 40%)
- implementing AI capabilities in database solutions (25 to 30%)
The exam is booked with Microsoft and remains optional. ITTA, a Microsoft Learning Partner, delivers this AI-enabled database training with MCT-certified trainers, in Geneva and Lausanne or in a virtual classroom, using official Microsoft courseware.
FAQ
What is the difference between the DP-800 and DP-300 courses?
DP-300 is for administrators: high availability, backups and monitoring of Azure SQL databases. DP-800 is for people who build: data modeling, advanced T-SQL, CI/CD and AI features. The two courses complement each other well within a data team.
DP-800 or DP-420: which course should I choose for an AI application?
It depends on your database. DP-420 builds AI applications on Azure Cosmos DB, a NoSQL database. DP-800 is rooted in SQL Server, Azure SQL and Fabric. If your business data already lives in SQL, DP-800 saves you a migration.
Do I need to know Python to follow the course?
No. AI features are handled from the database, in T-SQL: calling external models, generating embeddings, vector search and RAG. Some background on embeddings and vectors helps but is not essential.
What level of T-SQL do I need before starting?
You should be comfortable writing queries with joins and aggregations. The course then moves quickly to CTEs, window functions, JSON and regular expressions. If you are new to T-SQL, the DP-080 course is good preparation.
Why build RAG in SQL rather than with a separate vector database?
Keeping vectors next to the source data makes synchronization simpler, applies the same security rules and lets you combine relational filters with semantic search in one query. The course gives you the criteria to pick the right approach for your context.
Does the course apply to on-premises SQL Server as well as the cloud?
Yes. The curriculum covers SQL Server, Azure SQL and SQL databases in Microsoft Fabric. From the first module, you learn to choose the right platform and then apply the same techniques on each one.