Скачать или смотреть New Breakpoint: S5 Ep4 | Agents, Orchestration & MCP: Peeking Inside the AI Toolkit

Welcome to New Breakpoint!

In this episode, host Michelle Sandford for a deep dive into building intelligent applications using AI toolkits, Azure AI Foundry, and MCP tools. Discover how developers across Asia-Pacific are leveraging Microsoft technologies to create smarter, safer, and more scalable AI solutions.

🔍 What You'll Learn:
• How to build intelligent applications using generative AI
• Best practices for prompt and context engineering
• How to use Azure AI Foundry for model selection and observability
• The role of MCP tools in standardizing AI integrations
• Techniques for testing and evaluating AI applications
• Real-world demo: AI assistant for a fictional bike store

📚 Chapters:
00:07 - Welcome to New Breakpoint
00:45 - Meet Rakesh
01:03 - What’s inside the AI Toolkit?
02:27 - Building blocks of intelligent apps
03:52 - Language models and Azure AI Foundry
04:58 - Prompt engineering and context management
06:37 - Multi-agent orchestration and use cases
07:56 - Guardrails and safety in AI
08:45 - History of LLMs and context limitations
10:00 - Retrieval Augmented Generation (RAG)
12:04 - Memory types in AI apps
13:42 - Schema enforcement and tool definitions
15:07 - MCP tools vs. function calling
17:04 - Standardizing integrations with MCP
20:30 - MCP architecture and components
24:45 - Trusting MCP servers and registries
27:00 - Observability and evaluation strategies
30:00 - Pre-production and post-production testing
34:45 - Demo: Contoso Bike Store AI Assistant
38:00 - Architecture walkthrough
41:00 - Azure AI Foundry thread logs
42:59 - MCP server implementation in .NET
46:00 - VS Code integration and testing
49:00 - Model verbosity and token efficiency
50:12 - Final thoughts and future of intelligent apps
52:12 - Wrap-up and live Q&A

[eventID:26377]