ACKF logoACKF
Machine Learning

Amazon Kendra

View on GitHub

Amazon Kendra

Amazon Kendra is a managed enterprise search service that indexes and returns relevant results from organisational content using machine learning-based relevance and natural language understanding. It operates as a managed search index with connectors that ingest documents from supported data sources and an API that applications call to run queries. It typically fits into architectures as the search backend for knowledge portals, support systems, and internal document search interfaces.

🗂 Resource Category

Artificial Intelligence and Machine Learning

🧠 Exam Memory Hook

Think: "Enterprise content + natural language queries + connectors = Amazon Kendra"

📖 Ownership

Classification: AWS Managed Service

AWS responsibilities: AWS operates and patches the underlying service platform, control plane, and the infrastructure that hosts the indexing and query runtime; AWS manages service availability, scalability of the managed platform, and platform-level security controls. AWS provides built-in connectors, indexing, query APIs, and maintains the managed runtime used by the service. AWS does not own or manage the customer's source data stored outside the Kendra index unless a managed connector configuration transfers data into the index.

Customer responsibilities: The customer configures indices, access control policies, data source connectors, relevance tuning, query settings, and credentials for connectors; the customer secures source data, IAM policies, network configuration for VPC access, and monitors query and index usage through CloudWatch and service logs. The customer prepares and classifies content, defines FAQs and synonyms when used, and implements application-level authentication and authorization for search APIs.

Patching responsibilities: AWS patches and maintains the physical infrastructure, hypervisor, and the managed service runtime that runs indexing and query functionality. There is no customer-managed guest operating system for the Kendra managed service; any customer-managed compute (for example Lambda functions used to transform content) must be patched by the customer, including their operating system when applicable, runtimes, libraries, dependencies, and application code.

🏗 Typical Architecture

💡 Top 5 Features

  • Natural language query support with relevance ranking.
  • Prebuilt connectors to common enterprise data sources.
  • Configurable relevance tuning and document boosting.
  • Incremental indexing and document synchronization.
  • Search APIs with fine‑grained access control integration.

✅ Top 5 Use Cases

  • Enterprise knowledge base search for internal documentation and FAQs.
  • Support portal search to help agents and customers find relevant articles quickly.
  • Search across mixed content types (documents, HTML, databases) from multiple sources.
  • Product documentation search to surface relevant guides and how‑tos to users.
  • Compliance or eDiscovery preparation where indexed search across repositories is required.

🏗 Architecture Placement

Amazon Kendra sits as the managed search/indexing backend in an AWS architecture; client applications or web UIs send search queries to Kendra's query API and connectors or ETL processes ingest content from S3, databases, or other supported sources into the Kendra index. It can be configured to access data in a VPC for private sources and integrates with IAM for access control and CloudWatch for monitoring.

🎯 Commonly Used With

- Amazon S3
- Amazon RDS
- AWS Lambda
- AWS Identity and Access Management
- Amazon CloudWatch

🌍 5 Real-World Examples

  • A healthcare provider indexes clinical guidelines and policy documents so clinicians can ask natural language questions and retrieve relevant protocols at point of care.
  • A retail business indexes product manuals and internal manuals so support agents locate troubleshooting steps during customer calls.
  • A government agency consolidates public reports and internal briefings into a searchable index to support research and policy drafting.
  • An education platform indexes course materials and FAQs so students and instructors can find answers via the platform search box.
  • A financial services operations team indexes procedures and regulatory documents to speed investigations and audit preparation.

🎓 AWS Exam Clues

  • Choose Kendra when the requirement emphasises natural language relevance and enterprise connectors over simple keyword search.
  • Consider Kendra if you need built-in connectors to S3, databases, or other enterprise repositories with configurable indexing.
  • If the data must remain in a private VPC, verify Kendra’s VPC access and connector capabilities during design.
  • Kendra provides managed indexing and query APIs; the customer remains responsible for securing source data and IAM configuration.
  • For simple keyword search over a single object store, a lower‑cost search option may be appropriate instead of Kendra.

📝 Quick Revision

Amazon Kendra is a managed enterprise search service providing natural language and relevance-tuned search across multiple data sources. Use it when you need ML‑based relevance, connectors, and configurable ranking; key considerations are data ingestion, access controls, and monitoring/index maintenance.

🏷 Keywords

semantic search • enterprise search • data connectors • indexing • relevance tuning • FAQ extraction • document ingestion • incremental sync • access control • query API • VPC access • search ranking