Platforms · Services

Built on the platform you already run.

Catalyx lands, contextualizes and governs plant data on whichever platform you have chosen: Cognite Data Fusion, Databricks, Microsoft Fabric, AWS, Google Cloud, Snowflake or your own on-prem stack. The instruments run there natively. Nothing leaves your tenant.

Talk to a platform engineer Where we deploy →
PlatformsYours. No preferred vendor FoundationAxiom, published natively First instrument liveWithin 90 days DataStays in your tenant
Why the platform matters

A model is only as good as the record beneath it. We build the record on your platform, not beside it.

Operators have already chosen a platform. The gap is between the raw historian feed and a governed asset model the instruments can trust. We close that gap in the platform's own primitives, so your team can run it after we leave.

Where we deploy

Any platform.
One method.

Same agents, same evidence gates, same ninety days. Only the primitives change. Three platforms we know deepest are below; the rest are in reach.

P&ID · 23-D-401 ASSET GRAPH FITIPI ASSET TAGWORK ORDERDOCTAGTAG AGREEMENT 98.7% AXIOM → CDF DATA MODEL
01 · Contextualization-native

Cognite Data Fusion

Where the asset graph is the product. Axiom's agents populate the CDF data model, and the instruments read it back.

i.
Data model designIndustrial data models for tags, assets, work orders and documents, written to CDF's data-modeling API.
ii.
Contextualization pipelinesP&ID and document parsing, tag-to-asset matching, every link scored for cross-system agreement.
iii.
Instruments as Cognite FunctionsEmber, Envelope and Meridian run inside CDF. Results land in Charts and Industrial Canvas.
iv.
GovernanceAccess groups, data-set lineage and an audit trail on every derived value.
Runs Axiom · Ember · Envelope · Meridian
PI · SAP · DOCS UNITY CATALOG BRONZE · RAW SILVER · RESOLVED GOLD · ASSET MODEL MODEL LINEAGEACCESSAUDIT DELTA · 14,208 TAGS · MLFLOW AXIOM → UNITY CATALOG
02 · Lakehouse-native

Databricks

Where scale and ML live together. Historians stream into Delta, Unity Catalog holds the lineage, MLflow holds the models.

i.
Lakehouse landingBronze, silver and gold for PI, SAP and documents. Streaming tables for the historians, Auto Loader for the drops.
ii.
Governed asset modelAxiom publishes the verified asset model to Unity Catalog, lineage intact, shareable across workspaces.
iii.
Physics-informed MLModels trained on the plant's own history, tracked in MLflow, served as endpoints the instruments call.
iv.
Instruments as Databricks AppsEmber, Envelope and Meridian run as native apps and jobs, on your compute, inside your workspace.
Runs Axiom · Ember · Envelope · Meridian
PI · OPC UA ONE COPY · ONELAKE ONELAKE EVENTLAKEWAREPOWERML HOUSEHOUSEHOUSEBI TEAMS · COPILOT DIRECT LAKE · PURVIEW LINEAGE AXIOM → ONELAKE
03 · OneLake-native

Microsoft Fabric

Where the operator already lives in Microsoft 365. One copy of the data in OneLake, every workload reading from it.

i.
OneLake foundationLakehouse and warehouse items for plant data. One copy, shortcuts instead of duplicates.
ii.
Real-time telemetryPI and OPC UA streams through Eventstreams into Eventhouse. KQL for the second-by-second view.
iii.
Semantic models for the plantDirect Lake models that put Ember, Envelope and Meridian into Power BI, Teams and Copilot.
iv.
GovernancePurview lineage, sensitivity labels and workspace roles mapped to your site hierarchy.
Runs Axiom · Ember · Envelope · Meridian
Also in reach

The method is the constant. Open formats and the platform's own primitives, wherever your data already lives.

04

AWS

S3 and Glue for the record, SageMaker for the models, Lake Formation for lineage and access.

05

Google Cloud

BigQuery as the asset store, Dataplex for governance, Vertex AI for training and serving.

06

Snowflake

Iceberg tables for plant data, Snowpark for the pipelines, Horizon for lineage and policy.

07

Historians direct

AVEVA PI, Honeywell PHD, IP.21. The instruments read the historian when there is no lake yet.

08

On-prem and open source

Kubernetes, Postgres and TimescaleDB, Parquet on object storage, MLflow. Air-gapped when the site demands it.

09

Something else

If it holds a time series and an asset hierarchy, the agents can land on it. Ask, and we will map it.

What we do on the platform

Land. Contextualize.
Govern. Run.

Four services, delivered in the platform's own primitives. Nothing proprietary sits between you and your data.

Platform run · live trace● running
▸ land 3 historians · SAP PM · 2,140 documents
▸ contextualize 14,208 tags → 612 assets
▸ govern lineage · access · 41 flags raised
▸ publish asset model → your platform
▸ deploy CX Ember · unit F-501
✓ first instrument live · day 61
01

Land

Historians, SAP and documents into the platform's storage layer, with physical-range checks on arrival.

02

Contextualize

Axiom's agents resolve tags to assets and documents to both, scored for cross-system agreement.

03

Govern

Catalog, lineage and access mapped to your site hierarchy. Every derived value traces to source.

04

Run

Instruments deployed native to the platform, measured against your own baseline.

Sources we land
How it starts

One platform. Ninety days.
Then it's yours.

Each phase ends at an evidence gate. The team running it at the end is your team.

Weeks 0 – 2

Assess

  • Tenancy, residency and security posture agreed
  • Inventory of historians, SAP and document stores
  • Target data model and governance map
Gate: a landing plan your architects sign
Days 0 – 30

Land

  • Ingestion live in the platform's own primitives
  • Axiom run on one unit
  • Verified asset model published natively
Gate: cross-system agreement above threshold
Days 30 – 90

Prove

  • First instrument deployed native to the platform
  • Value measured against your baseline
  • Your engineers own the pipelines
Gate: value shown on your data
The platform charter

What every platform engagement commits to.

i.

Native or nothing

If the platform has a primitive for it, we use it. No sidecar databases, no shadow pipelines.

ii.

One copy of the truth

Data lands once. Everything else is a view, a shortcut or a lineage edge.

iii.

Portable by design

Data models, pipelines and models are yours, in open formats. Parquet, Delta, Iceberg, open data models. Move platforms and the work moves with you.

iv.

Handed over

Your engineers run the pipelines by day 90. We move to research partner.

Begin

Bring your platform.
Leave with [ a governed model ]

Tell us which platform and which unit. We'll show the asset model and the first instrument running on it within ninety days.

Talk to a platform engineer Request a briefing