Three planes, ten properties, one principle: customer keeps the data.
Eisberg is built so that on the day you decide we are not the right partner, your data is already in Apache Iceberg in your bucket — and any other engine on earth can read it. That is the whole point.
Control plane
Eisberg-managed
- API · Auth · Billing
- Agent scheduler & MCP gateway
- Tenant config · Policy plane
- Intelligence OS · Lineage graph
Compute plane
Neo-cloud GPUs
- GPU-accelerated query workers
- Distributed federated query workers
- Sensor-scale stream processing
- Per-query routing · Auto-fallback
Data plane
Customer-owned
- Apache Iceberg · Open format
- S3 / Azure Blob / GCS
- Customer-managed KMS keys
- Eisberg never stores customer data
What is structurally true about Eisberg.
Components are deliberately not enumerated here — implementation choices are part of the platform's competitive surface. What we will commit to publicly is the property each layer guarantees.
| Layer | Property |
|---|---|
| Storage | Apache Iceberg on customer-owned object storage. Open format, customer KMS keys, structurally portable. |
| Catalog | Open-standard catalog with namespace branching and time travel by reference. |
| Query path | Three engines selected per query for cost and latency: a sub-millisecond path for agents, a federated path for cross-source work, a GPU path for heavy aggregations. |
| Streaming | Sensor-scale ingestion, change-data-capture from operational stores, auto-classification on arrival. |
| Orchestration | Cloud-agnostic Kubernetes manifests across hyperscalers and neo-clouds. Same configuration everywhere. |
| Workflow | Durable execution model for pipelines and agentic flows — survives infrastructure failures and resumes. |
| Policy | One policy plane that every component reads from. Row-level, column-level, action-level, agent-level — all expressed as code. |
| Lineage | Open-standard lineage events at every layer. Cross-system chain of custody, replayable on demand. |
| AI router | Provider-agnostic LLM dispatch. Any model, routed to the cheapest capable. No hardcoded vendor lock-in. |
| Caching | Predictive cache aware of your data layout, with measured cache-hit accounting that informs routing. |
Detailed component-level architecture, threat model, and the full implementation rationale are available under NDA. Most serious enterprise buyers want to see them; we share them with design partners and prospects we are working a real opportunity with.
Built for the next hardware shift, not the last one.
Storage, caching, query routing, and compute scheduling are written behind interfaces so the next generation of memory and compute hardware slots in as a new backend — without touching application code. Incumbents have to rewrite themselves to make the same move.
Detailed forward-architecture rationale is available under NDA. The short version: every layer that hardware will reshape was written behind an interface so the day-one cost of switching backends is bounded.
Want the full architecture deck?
The deep-dive — including component choices, threat model, agent governance specification, and forward-architecture rationale — is available under NDA. We share it with design partners and serious prospects on the second call.