INTERNAL · LIVE BY MONTH 12
Data-Centre Ops KG
The operating graph of Cithorum's 1 PB Jam Proof Cloud. Customers, buckets, workloads, nodes, incidents, restores, capacity, energy, cost — all on top of live Jam telemetry.
KNOWLEDGE GRAPH
Turn data into intelligence.
Storage is one thing. Understanding what your data means is another. The Cithorum KG sits on Jam telemetry as the software intelligence layer — operating dashboards first, vertical SaaS next.
01 — WHY A KG
Compression saves money. Knowing what data means saves time.
Compression saves money. Indexing makes data findable. Knowing what each entity means — and how it relates to every other entity — is what turns infrastructure into intelligence. The first two are table-stakes. The third is the product.
The KG sits on Jam telemetry. Every TB Jam processes generates structured signal: who put it there, what it cost, when it was last restored, how it relates to other workloads, which customer owns it, which vertical it belongs to. Every restore event, every dedupe match, every quota change becomes a typed edge. The data centre is the first knowledge graph.
That is why we run the KG internally first — as Data-Centre Ops KG, the operating graph of Cithorum's own 1 PB proof environment. It graphs every customer, bucket, workload, node, incident, restore, capacity headroom, energy draw, and cost line on top of live Jam telemetry. It is the first proof that the substrate works, and it is the artefact every external vertical inherits.
Once that substrate is written down, the same core seeds every vertical KG product. Workspace, Law, Medical Sequencing — each is a different schema and a different buyer, but the entities, the provenance model, the permissions, and the AI tooling are shared. One core. Many verticals.
02 — WHAT IT CAN DO
One core. Four verticals. Same entities, different schema.
Click a vertical. Watch which entities it reads, which it writes, and how the graph re-wires the relationships in real time. One core, many products — same provenance, same permissions, same audit trail.
03 — HOW IT STACKS
Customer data → Jam → telemetry → KG → verticals.
Five layers, one direction of dependency. Each layer adds structure to the one below. Verticals inherit everything underneath.
DATA
Customer datasets
Raw bytes — VM snapshots, FASTQ, telemetry, documents, sensor archives. Customer-owned, customer-controlled, customer-located.
INFRA
Jam infrastructure — compression + restore
The data plane. Compress, dedupe, index, envelope, reconstruct. Per-customer isolated buckets, deterministic restore, hash-verified bytes.
SIGNAL
Telemetry & metering
Every operation generates structured signal: actor, dataset, ratio, latency, restore-event, cost. Streamed into the KG core in near real time.
CORE
KG Core — entities, relationships, provenance
Typed nodes, typed edges, schema versioning, per-edge permissions, full provenance chain. The shared substrate every vertical inherits.
FIRST VERTICAL
Data-Centre Ops KG — Cithorum's own proof environment
Graphs every customer, bucket, workload, node, incident, restore, capacity headroom, energy draw, and cost line of the 1 PB Jam Proof Cloud. Internally consumed first; sold as a stand-alone product to data-centre operators second.
EXTERNAL VERTICALS
Vertical KG products
Medical KG · DESIGN-PARTNER PIPELINELaw KG · DESIGN-PARTNER GATEDWorkspace KG · BETA
04 — THE VERTICALS
Internal first. External one at a time.
Data-Centre Ops KG runs on our own 1 PB proof environment. Every external vertical inherits its substrate.
DESIGN-PARTNER PIPELINE · BIRAC-TRACKED
Medical KG
Multi-modal medical data infrastructure — imaging (PACS), sequencing (FASTQ), EHR, device tracking, clinical trials. Compressed at the data plane, unified into one audit-trailed graph. HIPAA-aligned, DPDP-ready, sovereign by default.
Read more →DESIGN-PARTNER GATED
Law KG
Matter memory, precedent and evidence graph, private AI over legal documents — without sending a single matter to a public model.
Read more →BETA · DESIGN-PARTNER GATED
Workspace KG
One map of your business across finance, customers, people, projects, communications, and documents.
Read more →05 — THE SHARED CORE
What every vertical inherits.
Build the substrate once; the verticals are then a schema, a pricing surface, and a buyer.
Entities + relationships
Typed nodes, typed edges, schema versioning. Add a new entity type without breaking the live graph or the dependent verticals.
Provenance
Every fact traces to a source event with timestamp + actor. No ungrounded assertions. Auditable end-to-end.
Permissions
Per-tenant, per-edge, per-attribute access control. A vertical can expose the graph to AI without leaking the underlying records.
Workflows + AI
Query primitives, an AI assistant over the graph, and an audit trail of every model call. The AI sits on the graph, not on raw documents.
06 — DESIGN PARTNERS
Paid design partners. One vertical at a time.
We launch one vertical at a time. Each starts with a paid design partner: six months, signed scope, exclusive vertical-launch access, named technical contact. We do not fund three verticals from general burn — each has to earn its launch from a customer who has skin in the outcome.
Talk to us about a design partnership →