Protein research copilot using Amazon Bedrock AgentCore

Use case typeTherapeutics researchUpdated Jun 23, 2026

Protein researchers face a time-consuming challenge: manually searching through thousands of peptide sequences to find structurally similar candidates is slow, error-prone, and requires deep domain expertise to interpret results. A conversational protein research copilot combines natural-language query parsing, protein embedding generation, vector similarity search, and AI-generated scientific summaries in a single interface.

Organization
Henry Schein One
Industry
Healthcare
Location
United States
Published
June 2026

Reported outcomes

−33%

gpu fleet reductionOther quantified impact

10,000 locationsactive locations11,000,000 x-raysx-rays processed1,500,000 x-rays per weekweekly x-ray volume1.4 secondsmedian latency2.2 secondsp90 latency

Strategic outcomes

Cost efficiencyReduced manual peptide search to a single conversational workflowSpeed & agilityReturned ranked and summarized research results quicklyScale & capacityValidated a global rollout path for 40,000 locations across four regionsCustomer experience & trustEnabled point-of-capture image feedback for clinicians and techniciansOther strategic outcomeImproved imaging quality feedback and claim readiness in the workflow

Primary read

Use case focus

Showing 2 of 2

  • 1Therapeutics research
  • 2Semantic search
  • Manual peptide search is slow, error-prone, and requires deep domain expertise.
  • Researchers need to find structurally similar candidates across large biological datasets more efficiently.
  • Built a conversational protein research assistant using the Strands Agents SDK to orchestrate three specialized tools in one runtime.
  • Used a parser tool to extract structured search parameters from natural-language queries.
  • Generated protein embeddings with an Amazon SageMaker AI serverless endpoint running ESM-C 300M.
  • Stored peptide embeddings in Amazon Aurora PostgreSQL-Compatible Edition with pgvector for cosine similarity search and metadata filtering.
  • Deployed the orchestrator to Amazon Bedrock AgentCore and the frontend to AWS Fargate, with AWS CodeBuild and infrastructure as code supporting deployment.
Technologies
Amazon Bedrock AgentCoreAmazon SageMaker AIAmazon Aurora PostgreSQL-Compatible EditionAmazon Elastic Kubernetes ServiceAWS Cloud WANAmazon FargateAWS CodeBuildpgvectorStrands Agents SDK
  • Reduced a multi-step researcher workflow to a single natural-language query interface.
  • Returned ranked results and scientific summaries in under a minute, with longer latency on cold start.
  • Demonstrated a cost-efficient serverless deployment pattern for intermittent research workloads.
Architecture

A Streamlit frontend runs on AWS Fargate and sends queries to a Strands agent in Amazon Bedrock AgentCore. The agent orchestrates a parser tool, a search tool, and a summarizer tool. The search tool calls an Amazon SageMaker AI serverless endpoint hosting ESM-C 300M to generate embeddings, then searches embeddings stored in Amazon Aurora PostgreSQL-Compatible Edition with pgvector via the Amazon RDS Data API. AWS CodeBuild is used for container build and deployment, and AWS CloudFormation provisions the infrastructure.

Sources & evidence1
Groundedness: 4/5Type: Blog PostPublished: Jul 10, 2026Publisher: AWSEvidence: VendorConfidence: Medium
Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases
Customer evidence
More from Henry Schein One
Provider evidence
AWS provider pageAWS ranked cases
Market context
Healthcare AI use casesHealthcare industry insightsUnited States country insights
AI patterns
AgentMulti-agent
Browse the catalog
All AI use casesAI use-case typesTherapeutics research use cases

Was this useful?