AI × AEC · CURATED INSIGHT
Octave Builds Multi-Tenant AI Platform on Amazon Bedrock
Octave deployed a multi-tenant AI architecture on Amazon Bedrock using Cohere Embed v4 and Anthropic Claude. The system isolates data across tenants and accelerates document triage.
Octave built a multi-tenant AI foundation on Amazon Bedrock for its Reliance quality management platform. The architecture uses Cohere Embed v4 to ingest multimodal documents into per-tenant Amazon OpenSearch indexes and isolates user sessions using dedicated microVMs in AgentCore runtime. The Complaint Triage Assistant applies Anthropic Claude models on Bedrock to summarize incoming records and match past corrective actions.
Key takeaways
- Octave deployed Cohere Embed v4 on Amazon Bedrock, achieving 1.4 million ingested documents per 24 hours per tenant through batch packing in 128K context windows.
- Tenant isolation is maintained by running sessions in dedicated microVMs, while Bedrock Guardrails enforce input and output content filtering to block prompt injection.
Source：AWS Machine Learning Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human.