ChatBot Learning Assistant with Large Language Models
summary
Built a production MVP in five months: a learning chatbot that answers course questions with retrieval-augmented generation, returns chapter references and illustrations, and offers auto-generated quizzes and flashcards. Uses OpenAI embeddings (ada), GPT-3.5/4 for prompts and citation, then replaces some components with GTS small locally for intent classification; stores convo state in Redis, Postgres and runs on Kubernetes. Latency cut by moving quiz/flashcard generation to background; evaluation remains manual.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:51Why build a custom chatbot
- 7:24Traditional mentor problems
- 10:41Building the chatbot with large language models
- 13:10UI and knowledge base
- 15:41Intent classification
- 19:24Socratic dialogue mode
- 21:11Dynamic quizzes and flashcards
- 25:27Optimizing quiz latency
- 28:40Lessons learned
- 30:28Q&A
related talks
25mLets Build A Node.js ChatGPT-Like Bot With AI APIs
51mA ChatGPT on your own data: how does it work?
25mFrom Treatise to Tech: The Development of ChatSOC, an AI Research Assistant
39mBuilding Your Own Bot with GPT: Harnessing the Power of AI...
29mA Lean Journey of AI Chatbots in the Post-ChatGPT Era
31mMachine Learning-Enabled Chatbots
01h 00mBuilding an Expert Q&A Bot with Open Source Tools and LLMs
51mChatGPT as a developer's personal servant