MSD text-to-SQL for healthcare analytics using Amazon Bedrock
MSD collaborated with AWS Generative AI Innovation Center to build a text-to-SQL generative AI solution for complex healthcare databases. The system helps analysts and data scientists turn natural-language questions into executable SQL, reducing manual query writing and improving access to data for decision-making.
- Organization
- MSD
- Industry
- Pharma
- Location
- United States
- Published
- November 2024
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- MSD
- Provider
- AWS
- Maturity
- Unknown
- Linked source
- AWS Machine Learning Blog
No explicit deployment-stage evidence found.
Primary read
Use case focus
Showing 3 of 3
- 1Text-to-SQL
- 2Knowledge Retrieval
- 3Data Access Automation
- MSD implemented a custom text-to-SQL pipeline using Anthropic Claude 3.5 Sonnet on Amazon Bedrock and the Bedrock Converse API.
- The solution uses detailed schema and column descriptions in prompts, lookup tools for coded values such as sex, race, and state, and a tool-calling loop to resolve codes before generating SQL.
- The workflow also rewrites long diagnosis-code lists with placeholders and asks the model to produce an interpretation of the question before the SQL to reduce ambiguity.
Architecture
A custom text-to-SQL pipeline sends a system prompt with schema, table and column descriptions, few-shot examples, and user input to Anthropic Claude 3.5 Sonnet on Amazon Bedrock via the Bedrock Converse API. If the model requests a lookup tool, the pipeline retrieves coded values such as sex, race, or state and appends the result back into the prompt before continuing; if the model produces SQL, the application returns the query and explanation.
Implementation partners1
Sources & evidence1
- Customer explicitly identified
- Technical implementation details available
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?
Community
Comments
No published comments yet.