Code First LLMOps With Azure AI & Prompty
summary
A code-first workflow demonstrates a working recipe generator built with Azure AI, Prompty and Semantic Kernel (also exportable to LangChain), producing valid JSON from prompts and running locally with F5; Prompty keeps prompt, parameters and model together so each prompt can pick its own deployment (including free GitHub models), tests use PromptFlow, and observability via OpenTelemetry traces and Application Insights shows token usage and costs and a live outage fixed by switching to GPT-40 mini.
AI-written summary · about
chapters (11)
- 0:00🚀 Introduction: LLMOps and the developer's mindset
- 4:00Proof of concept: Starting with minimal effort to prove the idea
- 8:00Tooling overview: From prompt engineering to deployment
- 12:00The challenges of model selection and how it affects development
- 16:00Demo: Building a simple recipe generator using Azure AI
- 20:00The role of Prompty in improving prompt engineering
- 24:00Deployment and monitoring with Azure OpenAI
- 28:00Observability: Tracking token usage, errors, and costs
- 32:00Best practices for efficient model use and cost management
- 36:00Using semantic kernel and orchestrators to manage LLMs
- 40:00Closing thoughts and invitation to connect
related talks
40mCode-First LLMOps from prototype to production with GenAI tools
43mBuild and Maintain your Company Copilot with Azure ML and GPT-4 Q&A
46mPractical End-to-End AI Development using Prompty and AI Studio
01h 08mIn Prompts We Trust
27mBuild a RAG-based Retail Copilot Code-First on Azure AI Foundry
46mBuild and maintain your company Copilot with Azure ML and GPT-4
24mLLM App Development Using PromptFlow and Azure Cosmos DB
01h 58mBuild, Evaluate and Deploy a RAG-Based Retail Copilot with Azure AI