How IskayPet unified corporate knowledge with Amazon Bedrock Managed Knowledge Bases
IskayPet built a multilingual corporate knowledge assistant using Amazon Bedrock Managed Knowledge Bases and Claude Haiku 4.5, eliminating the need to operate their own vector stores, embedding pipelines, or ingestion jobs. The solution leverages Bedrock’s native SharePoint connector for direct ingestion and its built‑in hybrid search (semantic + BM25) to retrieve accurate answers across four languages and three countries. Engineers can now deploy a production‑grade RAG system that scales without custom infrastructure while providing source citations for trust.
⚡ Key Takeaways
- Managed RAG infrastructure provided by Amazon Bedrock Managed Knowledge Bases removes the need for self‑hosted vector databases and embedding pipelines.
- Native SharePoint connector ingests documents directly from existing SharePoint repositories, avoiding data migration.
- Built‑in hybrid search combines semantic retrieval with BM25 keyword search to capture exact terms like product codes.
- Claude Haiku 4.5 is used as the LLM for answer generation within the Bedrock Knowledge Base workflow.
- The system routes queries to the appropriate knowledge domain (area, country) to prevent cross‑contamination of results.
By offloading RAG plumbing to Bedrock, engineering teams can rapidly launch trustworthy, multilingual knowledge assistants that scale with corporate growth while reducing operational overhead and cost.
✅ Practical Steps
- Configure the Amazon Bedrock Managed Knowledge Base to use the native SharePoint connector for ingesting your existing SharePoint documents.
- Enable the built‑in hybrid search option to combine semantic and BM25 keyword retrieval for your knowledge base.
- Deploy Claude Haiku 4.5 within the Bedrock Knowledge Base pipeline to generate natural‑language answers and source citations.
Want the full story? Read the original article.
Read on AWS ML Blog ↗