Generative AI for the .NET Developer
summary
A .NET app wired to Azure OpenAI and a Postgres + PGVector store indexes ~1,000 podcast episodes (embedding vectors of ~1,536 dims) — indexing takes about a day and returns 83% relevancy on matches; it uses GPT-3.5-turbo for streaming token completions, function/tool JSON outputs to extract summaries/topics, and a background worker + ASP.NET Core frontend; text is turned into embeddings for semantic search (RAG pattern) and into speech via a custom neural voice trained on 34 utterances — prototype works end to end but requires prompt grounding and chunking to stay accurate.
AI-written summary · about
chapters (11)
- 0:00Generative AI for .NET
- 1:46Podcast chatbot demo
- 6:00How LLMs generate text
- 13:25Prompt engineering basics
- 20:05Text generation with .NET and Azure OpenAI
- 27:55Embeddings and vector databases
- 31:26Function calling in LLMs
- 36:00Speech processing with AI
- 41:27RAG patterns and model knowledge
- 44:39Podcast chatbot architecture
- 55:14Q&A: Training vs grounding
related talks
48mGetting started with generative AI using Azure OpenAI Service
01h 10mPractical Applications of Generative AI: How to Sprinkle a Little AI in Your App
01h 01mUnlocking the Potential of AI
24mInfuse Generative AI in your apps using
46mPower AI apps and develop rich experiences with Azure SQL Database
44mSupercharge your data with Azure AI Search and OpenAI
59mOrchestrate your AI with Semantic Kernel
12mBuilding Generative AI Apps with Azure Cosmos DB