One logical lake for analytics, RAG knowledge, and agentic AI.
Microsoft OneLake is the unified data lake for Fabric: a single logical lake with shortcuts to ADLS, Amazon S3, and more. Gensten designs OneLake layouts for analytics and AI - bronze/silver/gold zones, document corpora for RAG, ACL strategies, and secure paths into Azure AI Foundry and Fabric IQ - so your agentic stack reads from governed storage, not shadow copies.
How OneLake fits into a production agentic stack.
Raw files, CDC, and partner drops via shortcuts.
Cleaned tables and normalized document stores.
Business-ready datasets and RAG collections.
Embedding pipelines, Fabric IQ, Foundry tools.
We store source documents and derived chunks with lineage so every agent citation maps back to an authoritative object in OneLake - essential for audits and content refresh.
Shortcuts let Fabric IQ and agents use data where it already lives. We inventories sources, choose shortcut vs. copy patterns, and prevent “RAG islands” that drift from the lakehouse.
OneLake is Fabric’s lake, but it is also the right foundation for AI: document corpora, feature tables, and agent retrieval can all live (or shortcut) there under one governance model.
OneLake is the unified logical data lake for Microsoft Fabric - one lake for files and tables, with shortcuts to ADLS, S3, and other sources without endless copy sprawl.
Store authoritative documents and derived chunk/embedding datasets with lineage, then index into Azure AI Search or vector stores so agents cite lake-backed sources.
Shortcuts reference data in place (e.g., ADLS Gen2 or S3) inside OneLake so Fabric IQ and AI workloads can use existing lakes without duplicating every byte.
Bronze (raw), Silver (cleaned), Gold (business-ready / RAG collections), plus clear AI consumption paths. Sensitive domains get stricter workspaces and ACLs.
Yes. That coexistence is a major advantage: BI tables and RAG document corpora share governance, lineage, and domain ownership.
We inventory sources, prefer shortcuts over copies, version derived chunks, and require citations back to OneLake objects so shadow vector-only stores do not diverge.
Not necessarily. OneLake is the lake foundation; warehouses/lakehouses in Fabric still provide compute and serving. We design the right mix for analytics and agents.
Workspace RBAC, sensitivity labels, and retrieval-time ACL filters ensure Foundry agents only access authorized domains and documents.
Map domains and sources, define bronze/silver/gold and RAG collections, implement shortcuts, then connect embedding/index pipelines and Foundry retrieval tools.
Unify analytics and AI semantics on Microsoft Fabric IQ - lakehouse intelligence that powers agents, BI, and decision systems.
Production Retrieval-Augmented Generation: chunking, embeddings, vector search, reranking, citations, and agent-ready knowledge retrieval.
The foundation of RAG quality: intelligent data chunking, embedding model strategy, metadata design, and index hygiene.
Build, evaluate, and operate enterprise agents and models on Microsoft Azure AI Foundry - with governed catalogs, tools, and Foundry IQ intelligence patterns.
Talk to Gensten about ADLC, RAG, LLM building, Foundry IQ, Fabric IQ, and OneLake - scoped to your KPIs and compliance needs.