Conversations With AI: Transforming Conference Learning & Data Engagement
summary
A conference pipeline records session audio via a TazCam or Anker pucks, transcribes offline with Whisper, scrubs speaker IDs, builds a RAG and knowledge graph (Infernotus) and spins up an LLM chatbot in LM Studio or Hugging Face in about 30–60 minutes per conference day. It handles 16 hours of programming from day one, processes locally to avoid cloud leakage, and the live demo chatbot asks follow-ups, compares sessions, and works as a one‑hour proof‑of‑concept deploy that attendees can take home as a conference assistant.
AI-written summary · about
chapters (16)
- 0:00Introduction and setup
- 1:51Speaker background and goals
- 3:58AI for data conversation
- 7:28Tools for daily efficiency
- 10:15Conference data challenges
- 11:58Knowledge graph tool demo
- 12:42Recording equipment setup
- 13:24Offline transcription process
- 14:28Local AI model setup
- 15:30Language model parameters
- 16:54Building the conference chatbot
- 18:00Live demo of chatbot
- 19:52Chatbot features and reset
- 20:33Processing and future steps
- 21:15Knowledge graph insights
- 21:58Analytics and conclusion
related talks
17mMy AI Clone Attended Conferences While I Flew Planes - Tanisha Sharma
16mPlenary session
01h 02m[Full Workshop] Building Conversational AI Agents
10mHow to go from opening Visual Studio to interacting with a live chat bot in 10 minutes
35mThe Podcast Pipeline Nobody Sees: Behind the Scenes with AI Agents and AI gen
01h 01mLessons Learned Building the Ultimate AI Bug Reporter
24mConversational Interfaces for Chatbots and Artificial Intelligence
33mDesigning Voice Agents for Real Conversations