Build production-grade multi-agent capabilities with Microsoft Foundry

Microsoft via Microsoft Learn

Go to class Write review

Design advanced prompting strategies for production AI agents in Microsoft Foundry. Implement multiturn reasoning architectures with dynamic context injection, build layered prompt injection defenses, design system prompt frameworks with autonomy-level control, design multi-intervention guardrail architectures covering all four attack surfaces, implement prompt versioning and optimization workflows, and design fine-tuning strategies and data preparation pipelines.
By the end of this module, you'll be able to:
- Design multiturn prompt architectures with dynamic context injection for complex agent reasoning tasks
- Implement layered prompt injection prevention strategies for agents processing untrusted content
- Build system prompt frameworks that control agent persona, behavior boundaries, autonomy levels, and escalation behavior
- Design multi-intervention guardrail architectures covering input, tool-call, tool-response, and output surfaces
- Apply prompt versioning and optimization workflows to maintain and improve agent quality over time
- Design fine-tuning strategies and data preparation pipelines for domain-specialized agent models
Build enterprise-grade tool ecosystems for production multi-agent systems with the Model Context Protocol (MCP) and Microsoft Foundry. Develop custom MCP servers with authentication, implement dynamic tool selection and routing, build tool result validation with fallback pipelines, and govern tool ecosystems with dependency management.
By the end of this module, you're able to:
- Design and build custom MCP servers with authentication, production error handling, and observability
- Implement dynamic tool selection and routing logic that matches agent needs to available tools at runtime
- Build tool result validation and fallback pipelines for production reliability
- Govern tool ecosystems with dependency management and versioning strategies
Implement advanced retrieval-augmented generation (RAG) pipelines for production AI agents using Azure AI Search and Microsoft Foundry. Design hybrid search architectures, implement reranking strategies, configure dynamic knowledge source routing, and optimize chunking and embedding strategies.
By the end of this module, you're able to:
- Design hybrid search architectures combining keyword and semantic retrieval for high-precision knowledge access
- Implement reranking strategies to improve RAG retrieval quality and contextual relevance
- Configure dynamic knowledge source routing to direct queries intelligently across multiple knowledge bases
- Optimize chunking and embedding strategies for different content types and retrieval requirements
Design multi-agent memory architectures for production AI systems using Azure Cosmos DB. Implement semantic memory with vector storage, configure context window optimization strategies, design memory expiration and pruning policies, and build context-aware agent behaviors based on recalled history.
By the end of this module, you'll be able to:
- Design short-term and long-term memory architectures that match agent use case requirements
- Implement semantic memory systems using Azure Cosmos DB vector search for knowledge retention across sessions
- Configure context window optimization strategies that balance memory richness with cost and latency
- Design memory expiration, pruning, and consolidation policies for production deployments with compliance requirements

Syllabus

Design advanced prompting strategies for production AI agents
- Introduction
- Design multiturn reasoning prompt architectures
- Implement prompt injection defenses
- Build system prompt frameworks for agent control
- Design multi-intervention guardrail architectures
- Implement prompt versioning and optimization
- Automate prompt regression and optimization
- Design fine-tuning strategy and data pipelines
- Module assessment
- Summary
Build enterprise-grade tool ecosystems with MCP and Microsoft Foundry
- Introduction
- Design production MCP server architecture
- Build MCP servers with error handling and fallback
- Implement tool selection and routing logic
- Govern tool dependencies and versioning
- Module assessment
- Summary
Implement advanced RAG pipelines with Azure AI Search and Microsoft Foundry
- Introduction
- Design hybrid search architectures
- Implement reranking and context ranking
- Design dynamic knowledge source routing
- Optimize chunking and embedding strategies
- Module assessment
- Summary
Design multi-agent memory architectures with Azure Cosmos DB
- Introduction
- Examine memory architecture patterns
- Implement semantic memory with vector storage
- Optimize memory retrieval and context injection
- Configure context window optimization
- Design memory retention and consolidation
- Enforce memory privacy and audit compliance
- Module assessment
- Summary