Harnessing the Power of LLMs in Advanced Analytics
summary
Covers concrete LLM history and numbers (GPT2 1.5B parameters, GPT3 ~175B trained on ~3,000 years of text) and practical builds: early text-to-SQL at ~50–60% accuracy in 2019 versus QueryGPT saving ~140 hours/month at Uber and LinkedIn reaching ~95% accuracy; describes text-to-SQL, retrieval augmented generation variants (iterative, corrective, graph), prompt methods like chain of thoughts and tooling (langraph, langchain, length miss, length fuse), plus quantization and distillation tradeoffs (GPT40 vs GPT4 mini cost example).
AI-written summary · about
chapters (9)
- 0:00Introduction
- 2:43Attention is all you need
- 7:32Prompt engineering
- 10:18Agentic development
- 11:36Text to SQL
- 13:27Retrieval
- 15:17Agent Ops
- 17:57Scalability
- 19:14Conclusion
related talks
52mUnleash the Power of Large Language Models (LLMs)
35mLeveraging LLM for text augmentation in Topic Modeling
43mBeyond the Hype: A Realistic Look at Large Language Models
01h 02mA hacker's guide to open source LLMs
01h 00mUnleashGPT: Mastering the Implementation of Large Language Models
41mEnhancing LLMs with Graph Technology
27mProductionizing Open Source LLMs
37mExpanding the Capabilities of LLMs Through Community Collaboration