From Data to Precision: Fine-Tuning LLMs for Proprietary Query Languages
summary
Argues that fine-tuning plus a modern data foundation is required to turn natural language into exact, executable proprietary queries, not just LLM prompts. Recommends using Amazon Bedrock and new storage features like S3T tables for vector-scale retrieval, streaming via Kinesis/Kafka for real-time state, and fine-grain access tied to agent credentials. Claims concrete wins: Formula 1 cuts incident resolution from days to 20 minutes (86% reduction), Adobe builds Firefly in nine months, Canva saves $3.6M annually.
AI-written summary · about
chapters (12)
- 0:00Data strategy for generative AI
- 1:26AI agents evolution
- 4:20Data foundation in Agentic AI
- 6:35AI agents in onboarding
- 9:06Designing Agentic AI systems
- 11:59AWS freedom innovation trusted AI
- 15:15AI agents business use cases
- 17:26Trust security reliability in AI
- 18:35Innovation cost versus growth
- 20:47Organizational transformation with AI
- 22:53AWS scaling generative AI
- 23:58Q&A
related talks
30mThe Future of Data is Words
49mLLM inference and fine-tuning on secure enterprise data
20mData is Your Differentiator: Building Secure and Tailored AI Systems
40mLLMs & Data Warehouses: Reliable AI Agents for Business (Keynote)
01h 30mUsing LLMs to improve your Search Engine
15mFrom Words to Data: Using LLMs to Turn Questions into SQL
31mFine-tuning LLMs: A Cost-Benefit Analysis for Businesses
24mConversational Search with LLMs