Revolutionizing Customer Care: Our Journey with LLM and RAG into Production
summary
A two‑year build deploys an LLM + retrieval‑augmented generation stack into AT&T production with a secured vector store + semantic and hybrid search, chunking and reranking, prompt engineering and post‑processing, and CI/CD for scale. Agents see concrete gains — a cited case drops search time from 15 minutes to 50 seconds — and the project reaches nationwide launch, earns an AT&T service optimum award and a mention in Q2 earnings.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:25RAG and its importance
- 3:11Traditional keyword search challenges
- 5:41RAG implementation journey
- 7:47RAG pipeline components
- 10:33Language model parameter tuning
- 12:19Prompt engineering and post-processing
- 13:02RAG performance metrics
- 15:30Production challenges and trade-offs
- 17:37Access control and scalability
- 18:40Future enhancements and agentic workflows
- 20:08Summary
related talks
50mHow to Test LLM Agents
25mLLM Powered Agents with Kubernetes
32mHarnessing Large Language Models for Enhanced Customer Experience
48mGoing Beyond Prompt Engineering
48mUnlocking LLM Potential: From Exploration to Integration
14mThe Retail Revolution: Real-World Generative AI Applications
46mLeveraging AI and Automation in DevOps
21mTransforming search and discovery using LLMs