Building database queries with Amazon Bedrock
summary
Shows a working Text-to-SQL system that uses Amazon Bedrock plus vectorized schema embeddings (saved to OpenSearch/Aurora/RDS) and Glue crawlers to keep metadata fresh, then generates Athena SQL and validates it; failed queries are sent back with the error and schema snippet for automated refinement so queries become self-healing SQL. Demo includes a Jupyter notebook and a Bedrock agent wired to two Lambda tools; example fixes a misnamed column and returns correct results.
AI-written summary · about
chapters (22)
- 0:00Introduction to session
- 1:08Why text to SQL solutions
- 3:18Generative AI for SQL generation
- 5:05Challenges with LLM SQL solutions
- 6:31Building enterprise-grade text to SQL
- 7:15Four foundational blocks
- 8:00Amazon Bedrock as the core
- 8:40Schema discovery with AWS Glue
- 9:21Knowledge engineering process
- 10:45Vector embeddings and RAG
- 11:07Query intelligence with Bedrock
- 12:10Reliability engine for self-healing
- 13:35End-to-end robust solution
- 13:56Demo of self-healing SQL
- 15:46Alternate solution with agents
- 16:08Amazon Bedrock agents overview
- 19:41Agent solution architecture
- 20:23Fraud transactions data schema
- 22:28Bedrock agent setup details
- 24:36Demo of agent-based SQL query
- 26:22Text to SQL as a tech trend
- 27:05Resources and closing
related talks
31mLamRAG: From Data to Constructive Insights using Amazon Bedrock
24mBuilding GenAI applications with Amazon Bedrock
20mPower Amazon Bedrock applications with Neo4j knowledge graph
40mBuilding an AWS solutions architect agent with Amazon Bedrock
38mGenerative AI - Architectures and applications in depth
40mData Intelligence Platform on AWS
55mSecurely build generative AI apps & control data with Amazon Bedrock
20mData is Your Differentiator: Building Secure and Tailored AI Systems