AIVSE: Vector Search Essentials

Preparing SQL Server DBAs and Developers for AI-Powered Semantic Search

Available for 1-year access and lifetime access (no expiration)

Buy Now

Upon completion: students will receive a verifiable badge through Credly/YourAcclaim.com

Overview

Course Overview

This course prepares SQL Server DBAs and developers for AI-powered semantic search. You’ll learn how to store, index, and query vector embeddings directly in SQL Server 2025, enabling natural language search, recommendation systems, and retrieval-augmented generation (RAG) patterns.

The hands-on lab database and all demo scripts are available on GitHub. Whether you’re running SQL Server locally, on a VM, or in Azure, the core concepts and query patterns you’ll learn apply across all deployment options.

Instructor: Joe Sack

Level: 200 (intermediate)

Need Help Justifying Training? Here’s a letter to your manager explaining why SQLskills training is worthwhile and a list of community blog posts about our classes.

What You’ll Learn

  • Semantic search fundamentals and when vector search outperforms traditional full-text
  • Vector data types, embeddings, and how to generate them from text
  • Distance metrics (cosine, Euclidean, dot product) and when to use each
  • Approximate nearest neighbor (ANN) search with DiskANN indexes
  • Hybrid query patterns combining vector similarity with traditional filters
  • Production deployment strategies including performance tuning and monitoring

Platform Compatibility

The course content applies to vector search across the SQL Server family. Here’s what you need to know for each platform:

  • SQL Server 2025: Full support. The GitHub repo includes a .bak file you can restore directly to any local or VM-based instance.
  • Azure SQL Managed Instance: Full support when using the 2025 update policy. You can restore the .bak file directly.
  • Azure SQL Database: Supported with minor setup differences. Azure SQL Database doesn’t support .bak restores, so you’ll import the lab database via dacpac/bacpac or scripted deployment instead. Once set up, all queries and concepts work the same.

The way you write and reason about vector queries is effectively the same across all platforms. All theory around embeddings, schema design, RAG, and hybrid search is cloud-agnostic.

Curriculum

Module 1: Course Introduction

Sets the stage for the course, introducing the instructor and outlining what you’ll learn across all modules.

  • Course objectives and target audience
  • Instructor background and experience
  • Module overview and learning path
  • Sample database introduction (SemanticShoresDB)

Module 2: Understanding the AI Data Problem

Explores why traditional SQL search methods fail for semantic queries and introduces the four-part vector search solution.

  • The semantic search problem
  • SQL Server pre-2025 search toolbox (LIKE, full-text search)
  • SQL Server 2025 vector search features
  • Real-world property search demonstrations
  • DBA bridge: equality vs. similarity
  • The four-part vector search solution

Module 3: Vectors and Embeddings

Explains the building blocks of semantic search: how vectors represent meaning as numbers and how embeddings are generated.

  • What is a vector (arrays of numbers, dimensions)
  • Embeddings: converting text to numbers
  • Quantization and storage considerations
  • Working with embedding models
  • Tokens and document chunking strategies

Module 4: VECTOR_DISTANCE and Distance Metrics

Covers how similarity is measured between vectors using distance metrics, with focus on the VECTOR_DISTANCE function.

  • Distance metrics overview (cosine distance)
  • Why cosine distance for semantic search
  • VECTOR_DISTANCE function syntax and usage
  • Demonstrations: similar text produces similar vectors
  • End-to-end property search example
  • Full-text vs. vector search comparison

Module 5: Vector Search Fundamentals

Deep dive into approximate nearest neighbor search, the DiskANN algorithm, and vector index implementation in SQL Server 2025.

  • The performance problem (why naive search fails at scale)
  • k-NN search concepts
  • Approximate Nearest Neighbor (ANN) algorithms
  • DiskANN algorithm deep dive
  • Vector index implementation in SQL Server 2025
  • VECTOR_SEARCH function and syntax
  • Performance, accuracy, and monitoring

Module 6: Hybrid Search

Shows how to combine traditional SQL filtering with semantic similarity for real-world query patterns.

  • Why hybrid search matters
  • Pure vector search limitations
  • Hybrid query patterns with filters
  • Hybrid search with JOINs
  • Combining vector search with full-text search
  • Post-filtering behavior and considerations

Module 7: Production Readiness and Performance

Covers everything needed to deploy vector search in production environments, from capacity planning to security.

  • Embedding generation strategies (ETL, batch, in-database)
  • Embedding API performance and reliability
  • Capacity planning (storage and memory)
  • Monitoring with existing DMVs
  • Security considerations
  • Maintenance and schema changes
  • Testing and cost management
  • DBA roadmap for implementation

Ready to buy?

Questions?

If you have any questions not answered by our Immersion Events F.A.Q., please contact us.