My AI Clone Attended Conferences While I Flew Planes - Tanisha Sharma
summary
An AI clone attends a virtual conference while the creator flies at 10,000 ft, built as a four‑stage pipeline: LLMs, RAG, MCP with local streaming STT (“ears”), an LLM agent over a vectorized personal corpus (“brain”), TTS voice clone (“mouth”) and lightweight clip‑sync avatar (“face”). It runs fully autonomous with no rollback, handles a 15–20 minute live Q&A, and delivers prepared answers accurately; failures include hallucinations, audio race conditions that reply to the wrong question, and missing human‑in‑the‑loop safeguards despite an 80% streaming payoff.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:29Why send AI clone
- 3:18Four-stage architecture
- 6:56What worked well
- 8:45What broke badly
- 13:47Lessons and tips
- 15:57Human in the loop
related talks
22mAI First: Applications of the Future
22mConversations With AI: Transforming Conference Learning & Data Engagement
48mMastering the AI Shift: Adaptive Strategies for Software Evolution
01h 23mBuilding an AI Travel Agent That Never Hallucinates
01h 02m"This is fAIne". Panel discussion
43mGiving my wife a voice with Ruby and AI cloning (+ discussion)
01h 09mAI is eating the world (and Google)
01h 07mBuilding and Scaling an AI Agent Swarm of low latency real time voice bots