Ship code faster with AI-powered NoSQL schema design | DEM310
Marko Hotti and Sergiy Smyrnov demonstrate how GitHub Copilot and the Azure Cosmos DB Agent Toolkit can speed up NoSQL schema design, with AI-assisted schema generation, query optimization suggestions, and refactoring recommendations, plus local iteration using the new Mac/Linux Cosmos DB emulator.
Overview
The session focuses on accelerating Azure Cosmos DB NoSQL schema design using AI-assisted tooling.
Key points covered:
- Why NoSQL schema design is difficult in practice:
- Denormalization trade-offs
- Partition key selection
- Choosing appropriate data modeling patterns
- Using GitHub Copilot together with the Azure Cosmos DB Agent Toolkit to speed up iteration:
- AI-assisted schema generation
- Query optimization suggestions
- Refactoring recommendations
- Rapid local testing with the new Azure Cosmos DB emulator for Mac/Linux.
- A demo of schema evolution across three iterations, positioned as taking ~30 minutes versus days of manual design.
Session chapters (from the video description)
- Benefits of NoSQL databases like Azure Cosmos DB
- Real-world usage references: OpenAI, ServiceNow, and GitHub Copilot
- Demo setup and repository packaging overview
- Volume metrics and data scaling projections
- Optimizing the data model to prevent early mistakes and measure impact
- Emulator creating updated database containers with refined index policies
- Pre-run vs optimized results comparison
- Cost savings estimation from schema optimization
- Best practices for validation, automation, and continuous optimization