AI Chats: Challenges in Business Integration
summary
Shows runnable code using LangChain and vector databases to do sentiment tagging and RAG retrieval, measures retrieval with precision and recall on a golden set, and compares cloud API costs versus self‑hosted LLMs with back‑of‑envelope math. Demonstrates vector embeddings + chunking, reranking, and a hallucination detector (Nemo Guard R) that flags inconsistent answers (example: Polish Ekstraklasa goals returned 1800 800 700). Reports which pipelines scale, which reduce tokens and costs, and which setups actually work in production versus which fail noisily.
AI-written summary · about
chapters (16)
- 0:00AI chat challenges in business
- 1:04AI chatbots in production vs demos
- 3:14LLM business use cases
- 7:55Understanding randomness in large language models
- 10:02Frameworks for LLM apps
- 11:05Handling outdated knowledge in LLMs with external context
- 14:42Vector embeddings for retrieval
- 21:33Vector databases setup
- 22:16Measuring retrieval accuracy
- 25:51Improving retrieval & hallucinations
- 28:24Detecting LLM hallucinations
- 31:14LLM tools immaturity
- 32:37Open source vs cloud LLMs
- 34:22LLM integration costs
- 36:07Balancing AI risks & goals
- 36:50Question and answer session begins
related talks
25mLangChain with Azure AI: Building LLM powered Applications
21mChatGPT, LLMs, and LangChains - A Beginner's Guide
01h 13mBuilding Production-Ready AI Systems in 90 Minutes
01h 12mBuild Your Own AI Chatbot with OpenAI & LangChain
34mJavaScript-Powered AI: Crafting Intelligent Systems with LLMs and Vector Databases
41mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
01h 59mLLM applications with LangChain
40m4 pillars for building a modern AI application in JavaScript