HyperLake is a sovereign infrastructure platform that provisions AI agent infrastructure – including data, analytics, and governance – inside a customer's own cloud environment with zero compute markup.
What is HyperLake?
HyperLake is an agentic data cloud infrastructure platform that ingests data from OLTP databases (PostgreSQL, MySQL), cloud storage (S3, GCS, Azure Blob, R2), open formats (Iceberg, Delta, Hudi), streaming systems (Kafka, Kinesis), and 100+ SaaS APIs. It federates these sources through a unified SQL layer powered by Trino, applies real-time governance (RBAC/ABAC, column masking, row-level security), and exposes governed access to humans (analysts, scientists, engineers) and AI agents via SQL, REST APIs, and a vector database (pgVector, Qdrant, Milvus). The platform is deployed inside the customer’s own VPC, private cloud, or on-prem environment, and is built by HyperLake.
Key Features
- $0 compute markup — Cloud compute is billed directly by the customer’s cloud provider (AWS, GCP, Azure, OVH); HyperLake does not add a per-query or per-compute surcharge.
- Unified governance engine — A global policy layer evaluates every request (human or agent) with RBAC/ABAC, column masking, row-level filtering, and full audit trail version-tracking.
- Sovereign by default — Agents operate on data without moving it outside its secure environment; sensitive data remains under full owner control through confidential compute patterns.
- Immutable provenance — Every agent action, inference, query, and training run is recorded via immutable provenance logs for complete traceability.
- Data-as-a-Service APIs — Any SQL query can be exposed as a REST API with built-in authentication, consumable by agents without SQL knowledge.
- Aria AI Data Analyst (Beta) — An autonomous agent that chats with structured data, runs multi-step analytical workflows, and generates ML insights on demand.
- Apps close to data — Deploy custom apps, models, and functions (like Snowpark but open) directly inside HyperLake via curated Helm apps, running code where data lives.
- Write-back to lakehouse (Coming Soon) — Agents can persist enriched data, predictions, and analysis results back into governed Iceberg tables.
Who is it for?
- Platform teams building agentic infrastructure — They provision and manage a unified data and governance layer for multiple AI agents across departments, enforcing consistent access policies.
- Data engineers supporting autonomous analytics — They configure ingestion pipelines, set RBAC/ABAC rules, and expose governed SQL endpoints for agents that need to query federated sources without manual intervention.
- AI/ML engineers deploying retrieval-augmented generation (RAG) — They combine Iceberg tables with vector databases (pgVector, Qdrant, Milvus) to give agents grounded access to enterprise data for context-aware reasoning.
What can you do with HyperLake?
- Governed agentic analytics — Deploy AI agents that autonomously run multi-step analytical workflows (data discovery → insight generation) through a single SQL layer, with scoped RBAC/ABAC policies and full audit trails.
- RAG over structured + vector data — Build retrieval-augmented generation systems that query Iceberg tables alongside vector stores, grounding agent actions in real enterprise data without data movement.
- Self-serve data access for humans and agents — Analysts and AI agents share the same governed platform: humans use dashboards and reports, while agents consume the same data via REST APIs or SQL, with consistent access controls.
