See it work, then see exactly how it runs.
Run your own telemetry file through the live Engine, or walk a full site demo. Then scroll for the whole picture: how CalibrAIte connects every system like a puzzle, how dashboards are an add-on, how it runs on OpenShift, how data is stored and secured for a major site, how compliance is handled, and what it costs.
Live Engine upload
Drop a telemetry CSV, JSON or NDJSON file and watch every record move through the ten-stage deterministic pipeline: ingest, normalise, classify against WA regs, correlate, flag anomalies, route and audit. Same input, same output, every time.
Open upload demoLive site demoFull platform walkthrough
Walk a live site: events flowing across systems, the correlation panel linking them into one incident, and role-based dashboards for every person on site, from Control Room to HSE to GM. The Engine, plus the operating layer on top.
Launch live demoEvery system on site, snapped into one picture.
A mine site is a pile of systems that each see one part of the operation and are blind to the rest. Contractors run their own tools. Sites generate their own events. OEMs ship their own telematics. CalibrAIte is the piece that makes them all fit together, without replacing any of them.
Six steps, start to finish.
CalibrAIte deploys inside your own infrastructure. You keep control of the cluster, the data and the keys. Here is how it stands up, start to finish.
It runs on your OpenShift cluster.
CalibrAIte ships as a standard container image. It deploys into the Red Hat OpenShift cluster you already run, on your cloud account (AWS ROSA, Azure ARO) or in your own data centre, in your chosen region. Your platform team controls it the same way they control everything else on the cluster. No new infrastructure vendor to onboard.
Deployed with a single, governed manifest set.
The full set of OpenShift manifests is applied into a dedicated namespace: deployment, networking, secrets, autoscaling and an egress-locked network policy. It runs as a non-root, least-privilege workload behind TLS. It scales itself with demand and self-heals on failure. Your team applies it; your team owns it.
Your data lives in your own database.
The database runs inside your cluster on EDB Postgres for Kubernetes, with high availability, automatic failover, continuous backup, point-in-time recovery, full database audit logging and encryption in transit under your keys. Every event, document and record stays in your cluster, in your region. If CalibrAIte were ever switched off, your data remains entirely yours, in place.
Events flow through a Kafka backbone.
Your site systems feed an event stream built on Red Hat AMQ Streams, which is Kafka running in the same OpenShift cluster. Events are durable, partitioned by site and asset so each machine's history stays in order, and replayable for forensics and audit. The stream is designed to scale to tens of thousands of events per second, per site, and to buffer at the edge and replay on reconnect if a link drops.
From raw signal to cited, routed action.
Each event is normalised to one schema, classified against your own site documents and the WA regulations with the source cited, correlated against related events to surface developing patterns and cross-system impacts, then routed to the right role with the evidence and the recommended next step. Everything is read-only. CalibrAIte adds context and routes action; your people and your approved systems decide.
Integrated into your security and identity.
At deployment, CalibrAIte wires into your existing single sign-on, multi-factor authentication and access controls. Every data source connects through a scoped key with expiry and IP allow-listing, and network egress is locked down. CalibrAIte never sits on your OT network and never takes operational control; systems push approved signals to it, it reaches into nothing.
The bottom line: CalibrAIte runs in your cluster, on your Red Hat stack, with your data never leaving your walls. You get the intelligence layer; you keep full sovereignty over the environment it runs in.
Contractors
Drilling, blasting, haulage and maintenance contractors push their event signals into one endpoint. They keep their own systems; you get their data correlated with yours.
Sites
Fleet management, IVMS, fatigue and driver monitoring, collision avoidance, SCADA, weather and access control. Every site source feeds the same pipeline.
OEMs
Caterpillar, Hitachi, Komatsu, Epiroc, AVEVA, DJI and more. Existing telematics readers push straight to the API. No rip-and-replace of OEM contracts.
One push API. One universal schema. Every event classified against your own documents and the WA regulations, correlated across sources, and routed to the right role with a full audit trail. CalibrAIte never takes operational control. A competent person always decides.
The Engine does the work. Dashboards are the add-on.
A site buys the Engine first: the headless intelligence layer that ingests, classifies, correlates and routes. When teams want a surface to log into, the Platform dashboards attach on top, per node. Same intelligence layer underneath.
Engine
Headless APIData pipeline + AI classification + correlation + calibration. Output via API and webhooks into your existing systems. No UI required. CalibrAIte processes your site data and pushes intelligence out.
Platform (dashboards)
Add-onEverything in Engine, plus role-based dashboards for every person on site. Onboarding wizards, document management, contractor multi-tenancy, audit trails and workflows. 14 roles from Company Owner to Contractor Worker.
Priced per node. Dashboards on top. Setup once.
Transparent, node-based pricing that scales with the site. The customer hosts the infrastructure (OpenShift or cloud), so you only pay for the intelligence layer.
Engine (per node)
$7.50 / node / month
The core product. The headless intelligence layer every site buys first.
- Ingestion, normalisation and classification against your docs + WA regs
- Cross-system correlation, anomaly flagging and calibration
- API + webhook output into the tools you already run
- Priced per monitored node, billed monthly
Dashboards add-on
+ $4.00 / node / month
The Platform layer on top of the Engine. Optional, attach when teams want to log in.
- Role-based dashboards across the full site hierarchy (14 roles)
- Onboarding, document management, audit trails with diff views
- Contractor multi-tenancy and permissioned access
- Attach per node, on top of the Engine rate
On-site setup
One-time, scoped
Connector integration, deployment, document ingestion tuning and acceptance testing.
- Connector + telemetry integration for your machine types
- OpenShift / cloud deployment and security review
- Site document ingestion + severity-matrix tuning
- Vendor onboarding, auth wiring and acceptance testing
A node is any monitored thing that produces events.
A node is a single source CalibrAIte ingests, classifies and correlates: a haul truck, an excavator, a dozer, a light vehicle, a fixed plant asset, a SCADA tag, a proximity or fatigue unit, a drone, or a contractor system feeding the API. If it generates a signal we turn into intelligence, it is a node. You pay for what is actually monitored, and the bill scales up and down with the site, not a fixed licence you over-buy.
Mobile fleet
Haul trucks, excavators, dozers, loaders, light vehicles
Fixed plant
Crushers, conveyors, pumps, SCADA tags and historians
Safety systems
Proximity / collision avoidance, fatigue and driver monitoring
External feeds
Drones, contractor systems and OEM telematics readers
One rate per node for the Engine, plus the dashboards add-on per node if teams want the full Platform to log into. Dashboards are optional and attach only to the nodes a site wants surfaced.
Priced as a rounding error against what one stoppage costs.
CalibrAIte does the cross-system correlation a control room and HSE office do by hand today, across screens, after the fact. The per-node rate is set so it disappears next to the value it returns and the cost of the work it removes.
Less than the manual work it removes
A site already pays HSE and control-room staff to reconcile siloed systems by hand. Per node, CalibrAIte costs less than that recurring overtime.
A fraction of the OEM contracts it sits beside
Fleet, autonomy and telematics contracts run into the millions. CalibrAIte is a thin intelligence layer on top, not another platform to rip and replace.
Trivial next to one avoided stoppage
A single unplanned haul-fleet stoppage or lost-time injury dwarfs a year of per-node fees. Prevent one and it has paid for itself many times over.
And because the customer hosts the infrastructure, the per-node rate buys the intelligence layer itself, the classification, correlation, calibration and audit, not a markup on servers you are already paying for.
A reference 1,000-node site: about $1.16M saved a year.
Conservative model on a single 1,000-node site with 40 haul trucks. These are reference figures; a real Tier-1 fleet pushes every number up.
$800k
Downtime avoided
$191k
Compliance admin
$31k
Investigation speed
$75k
Safety (LTI avoided)
$61k
Maintenance
Runs in your infrastructure, on Red Hat OpenShift.
CalibrAIte ships as a containerised application with a complete OpenShift manifest set. Deploy it in your own cluster, in your own region, under your own control. Cloud, local cloud, or fully on-prem.
Containerised image
Three-stage build, ~180MB on Alpine. Published to a registry (Quay or your OpenShift internal registry) as calibraite:0.5.0 plus the engine image.
Full manifest set
Deployment, Service, Route, ConfigMap, Secret, NetworkPolicy, HPA, PVC, ImageStream, Namespace and BuildConfig. oc apply and it stands up.
Self-healing & scaling
Rolling updates with zero downtime, readiness and liveness probes on /api/health, and a Horizontal Pod Autoscaler from 2 to 8 replicas at 70% CPU.
Hardened by default
Runs as non-root (UID 1001), no privilege escalation, dropped capabilities, edge TLS with HTTP-to-HTTPS redirect, and an egress-locked NetworkPolicy.
We have already hosted CalibrAIte on OpenShift.
This is not a slide. We pushed both images to the registry, applied the OpenShift manifests into a live cluster, supplied the runtime secrets, and brought the application up end to end. The health endpoint reported healthy with the database connected, and the version endpoint confirmed all engine services wired. From there we pumped varied, messy API payloads into the live ingest endpoint and watched the Engine normalise them to one schema, classify each event against the WA regulations, correlate them, and flag the anomalies, all inside OpenShift. The exact path a Tier-1 site would take to production.
Images pushed
calibraite + engine to registry
Manifests applied
deploy/openshift into the cluster
Verified live
/api/health healthy, DB connected, ingest normalising
We never own your data. You do.
CalibrAIte is the processor, not the owner. The site is the controller. Your data is ingested, classified and stored on your behalf, inside infrastructure you control, and it is exportable or deletable on demand. We never resell it and never train shared models on it. Because CalibrAIte ships as a container, the database is a swappable dependency, so you choose where it lives.
Managed (Neon)
Serverless managed Postgres in the AWS Asia Pacific (Sydney) region, so data stays in Australia. The Neon project is on your account and keys; CalibrAIte connects through a scoped credential you can rotate or revoke. Fastest to stand up.
Best for pilots and smaller sites.
In your OpenShift (EDB / CloudNativePG)
Run Postgres entirely inside your own OpenShift cluster with the CloudNativePG operator, or its commercially supported build, EDB Postgres for Kubernetes, a Red Hat OpenShift alliance partner. The data never leaves your cluster.
Best for Tier-1 production and full sovereignty.
Your existing database
Already standardised on AWS RDS or Aurora, Azure Database for PostgreSQL, or on-prem Postgres? CalibrAIte just points at it. No new data-store vendor to onboard; it fits the DBA, backup and compliance regime you already audit.
Best when you have a database standard already.
Same containerised app for all three. Only the connection string changes, so you are never locked into our database, our cloud or our region.
Run the database in your cluster, on your terms.
With the in-cluster operator, the events, documents and embeddings all sit on your storage, in your region, under your keys. If you switched CalibrAIte off tomorrow, your data is still in your cluster.
High availability
Primary / standby with native streaming replication and automated failover
Encryption in transit
TLS by default, bring your own certificates or integrate cert-manager
Audit logging
PGAudit enabled declaratively for regulator-grade database audit trails
Point-in-time recovery
Continuous backup to object storage, restore to any timestamp
Cross-region DR
Replica clusters in other regions via WAL shipping or streaming
Vector search
pgvector in-cluster for cited document grounding (RAG)
Events & classifications
Every event, classification, correlation group, anomaly and audit trace is stored in Postgres. Reachable only over TLS through the egress policy.
Site documents
SOPs, MSMS, PMHMP and procedures are chunked and embedded with pgvector for cited retrieval. Used only to ground answers, never to train shared models.
Tenancy isolation
Every record carries a workspace_id with row-level security, so sites and contractors are isolated in a shared database. Enterprise and air-gapped deployments can move to a database per tenant.
No private data needed
Correlations work on equipment and event signals, not names or faces. PII is avoided by design and role-gated where it is unavoidable.
How an event is stored, and how events connect.
Every signal that enters the Engine becomes one structured event record in your database. Correlation is not magic and it is not a black box. It runs over those stored records with deterministic rules you can read off the screen. Here is exactly what is kept and how the links are drawn.
What one stored event record holds
Each event is persisted with its full chain of evidence, so it can be queried, correlated, exported and audited later. Nothing is thrown away.
Raw payload
The original message exactly as the source sent it, kept for audit.
Normalised event
Equipment id, equipment type, event type and metrics in one schema.
Classification
Class, severity band and the WA regulatory basis cited.
Anomaly
Whether it was flagged, and the literal reason it was flagged.
Correlation group
The id of any incident group this event was linked into.
Routing + audit
The action drafted, the role it went to, and the timestamped trail.
Stored against the site profile and the machine profile, indexed by time, so an asset's full history and a shift's full picture are both one query away. Exportable to PDF or CSV for a regulator at any time.
A developing fault on one machine
When two or more events land on the same equipment id inside the recent window, they are linked as one developing pattern on that asset, instead of being read as separate, unrelated alarms. A brake warning, then a derate, then an e-stop on the same truck become one story.
A cluster across different machines
When two or more events of the same classification appear across different assets, they are grouped as a fleet-wide cluster and flagged as a leading indicator. Three proximity events in one zone, or the same fault class on several machines, surface as one signal, not noise.
Both rules are deterministic: the same events always produce the same correlation, and the reason is shown on screen. On a live site the window is your stored event history, so correlation runs continuously across every source feeding the Engine, not just one upload. That is why a single record explains itself, but a stream is where the real value appears.
Why the correlations can't be hallucinated.
The fair worry with any AI on a safety system is that it invents a connection that was never there, an LLM confidently linking two events because they sound related. CalibrAIte is built so that cannot happen on the hot path. The link between events is never written by a language model. It is a deterministic match over the stored records, and it shows you the literal reason it grouped them.
Generate a plausible link
Feed two events to a language model and ask if they are related, and it will produce a fluent narrative either way. It can connect events that share nothing but vocabulary, and the same prompt can give different answers on different runs. That is a hallucinated correlation, and on a mine site it is dangerous.
Match on facts, then show the reason
Two events are only linked when they share a hard, recorded fact: the same equipment id, or the same classification inside the same time window. No model writes that link. It is set membership over your stored events, and the grouping reason is printed on the screen so a competent person can check it in seconds.
Deterministic, not generative
The correlation step is plain rules over stored fields. The same events always produce the same groups. No temperature, no sampling, no improvisation.
Grounded in recorded facts
A link requires a shared equipment id or a shared event class in a window. If that fact is not in the data, no link is drawn. It cannot invent a relationship that the records do not support.
Every link shows its reason
Each correlation carries the literal sentence that produced it: how many events, on which asset or class, in what window. Nothing is asserted without the basis next to it.
AI assists, never decides
Language models help read messy documents and draft wording. They never make the safe-or-unsafe call and never author the correlation. A person reviews and decides.
The same discipline runs through classification. Severity and regulatory basis come from explicit rules and thresholds against your documents and the WA regulations, each one cited, and each carries an evidence score rather than a claim of absolute truth. If CalibrAIte cannot point to the fact, it does not assert the link.
What it does today, and what it does next.
Straight about the line between what is built and proven and what is on the roadmap. The engine, the deterministic pipeline and the cross-system correlation are real today. The breadth of connectors and the formal Tier-1 assurances are where we are heading.
Built and proven today
- Push any messy payload to one endpoint and have it normalised to a single schema
- Deterministic classification with severity and a cited WA regulatory basis
- Cross-system correlation that groups related events into one incident, with the reason shown
- Anomaly flagging against operating bands, with the literal trigger recorded
- Full audit trail per event, exportable to PDF or CSV
- Runs in OpenShift in your own cluster; we have already stood it up end to end
- Live public upload demo and a full role-based dashboard walkthrough
On the roadmap next
- A paid pilot on a real, read-only site data chain to prove it on live operational data
- Correlation breadth beyond operations and safety, into maintenance, HR and finance overlays
- The formal Tier-1 assurance path: SOC 2, cyber and PI cover, OpenShift certification
- Identity-layer integration with the site SSO, MFA and runtime access controls
- Self-serve push API so any contractor or OEM can onboard a new source
- Hardware-agnostic connector catalogue, then in-house connectors and national fitting
The mechanism that matters, ingest, classify, correlate without hallucinating, is built today. What is ahead is breadth across more data sources and the formal assurances a Tier-1 site signs off before production.
Built to slot into a Tier-1 security posture.
Push-based and read-only by design. CalibrAIte never needs access to your OT network and never takes operational control. That makes integrating the security a major site demands straightforward, not a fight.
No OT network access
Systems push approved event signals to one endpoint. CalibrAIte reaches into nothing and cannot touch controls.
Scoped API keys
Per-source keys with expiry, rate limits, IP allow-lists and granular permissions. Revoke a contractor in one click.
Identity integration
Wire into the site identity provider and runtime access layer (SSO, MFA, machine and service identities) at deployment.
Role-gated access
Sensitive data is gated by role. Workers see what workers need; auditors get evidence; no surveillance by default.
Audit trail on everything
Every event and classification carries an immutable trace. Regulator-ready evidence, not a reconstruction.
Runs in your cluster
Your OpenShift, your secrets, your network policy. Data sovereignty is the default, not an upgrade.
Evidence, not verdicts. A human always decides.
CalibrAIte classifies every event against your documents and the WA mining regulations, cites the source, and routes a draft to the right role. It is a compliance evidence layer, not an automated authority over a person.
Cited against WA regs + site docs
Each classification points to the clause and the procedure it is based on. Auditors follow the trail; they do not take our word for it.
Draft-only, human-in-the-loop
Outputs are drafts with review disclaimers. CalibrAIte never makes a safe-or-unsafe call and never has final authority over a person.
Evidence score, not a truth score
We surface how well-supported a classification is, not a claim of absolute truth. The competent person makes the operational decision.
Regulator-ready packs
Export incident timelines, correlated events and citations as PDF or CSV evidence packs whenever a regulator or internal audit asks.
Where this goes next.
- 1
Pilot
A fixed-scope paid pilot on one real, read-only data chain. Prove ingestion, classification and correlation on the site's own data and documents.
- 2
Hardening
OpenShift certification path, identity-layer integration, and the formal security review a Tier-1 site requires (cyber, PI and the SOC 2 path).
- 3
Scale across sites
Open the push API to any contractor or OEM. One endpoint, one schema, every machine type. Self-serve onboarding for new data sources.
- 4
Connector supply
Hardware-agnostic connector catalogue, then in-house connectors and fitting, so even legacy machines can join the picture nationally.
What buyers and partners ask first.
The questions that come up in every serious evaluation, answered straight.
We already run MineStar, LinkOA or AVEVA PI. Why CalibrAIte?
We do not replace any of them. Each of those systems is excellent at its own silo and blind to the others. CalibrAIte sits on top and correlates across them, the cross-system stitching your control room and HSE office do manually today, after the fact. You keep every system you already paid for.
Does it touch our autonomy or control systems?
No. CalibrAIte is push-based and read-only. Systems send approved event signals to one endpoint; it reaches into nothing and never takes operational control. That is by design, so it is safe to run alongside autonomous fleets and critical controls.
How accurate is the classification, and who decides?
The hot path is deterministic, so the same input always produces the same result. Every classification is cited against your own documents and the WA regulations and carries an evidence score, not a claim of absolute truth. It is draft-only. A competent person always reviews and decides.
What about false positives and alarm fatigue?
Correlation reduces noise rather than adding to it: related events are grouped into one incident instead of firing as separate alarms. Calibration learns what normal looks like for your site over time, so anomalies stand out.
Where does our data live, and can we keep it in Australia?
Deploy the cluster and database in an Australian region. The route uses a .au host with an Australian certificate and egress is restricted to approved services only. The data is yours, in your boundary, by default.
How do you handle security for a major site?
No OT network access, scoped per-source API keys with expiry and IP allow-lists, role-gated access to sensitive data, an immutable audit trail on every event, and deployment into your own cluster with your secrets and network policy. We integrate with your identity provider and runtime access layer at deployment. We are building toward the formal assurances a Tier-1 site expects, including the SOC 2 path, and are upfront about where we are on that journey.
You are a new company. What is the continuity risk?
CalibrAIte runs in your OpenShift cluster, not locked to our cloud, and the data is yours and exportable. That means you are never stranded if circumstances change. CalibrAIte Pty Ltd is registered, and we are working through the partner, certification and insurance path a major site requires.
How is a node counted for billing?
A node is an actively monitored source, a machine, a fixed asset, a SCADA tag, a safety unit, a drone or a contractor feed. You pay for what is monitored in a given month, so the bill scales with the site instead of a fixed licence you over-buy. The exact counting method is agreed in the contract.
Want it running on your data?
Scope a fixed paid pilot. Your connectors, your documents, your cluster. We prove the intelligence layer on real site data with skin in the game on both sides.