Maxis
Maxis IntelligenceOracle Autonomous AI Database · on Exadata

One converged database.
Seven business domains.
Every data model.

Relational, JSON documents, AI vectors, spatial, graph, in-database machine learning and natural-language Select AI — all on the same data, in one Oracle Autonomous AI Database, in-region. One engine on Exadata in place of six stitched-together services: one security model, one audit trail, one PDPA-aligned boundary.

1
Autonomous Database
7
Business domains
8
Data models, one engine
1
In-region boundary
RelationalJSONVectorSpatialGraphIn-DB MLSelect AISecurity
Click any domain to dive straight in — or open the converged overview.
Synthetic demonstration data on Oracle Autonomous AI Database 23ai (Exadata). This is a capability demonstration; all figures are illustrative and are not Maxis production data.

Converged Overview

One database. Every data model. On Exadata.
🛡 Data stays in-region · Autonomous AI Database

One Converged Database. On Exadata.

This entire application runs on a single Oracle Autonomous AI Database. Relational, JSON documents, AI vectors, text, spatial, graph, in-database machine learning and natural-language Select AI — all on the same data, in one engine, in-region. For a Microsoft-heavy estate that means retiring a warehouse + Cosmos + AI Search + a vector DB + a graph DB + an external ML platform, and the data movement, extra licences and patch surface that come with them.

Live data-model inventory — one MAXIS schema, queried live on this page

Each tile is a different data model. The "replaces" line maps to the Microsoft-stack equivalent Maxis would otherwise run.

🛡 Sovereign by design

Every data model lives in one Autonomous Database, in-region. Nothing is copied to a separate warehouse, vector store or geo service — one place to secure, one boundary keeping subscriber data inside Malaysia. Aligned with PDPA 2024.

💰 Less sprawl, less cost

One engine on Exadata replaces six services. Fewer licences, fewer integrations, fewer copies of customer data, a smaller patch surface — and one security model instead of six.

🧠 AI-ready on live data

Vector search, Select AI and in-database ML run on the live operational tables — no pipelines shuttling subscriber data to an external AI stack. The data and the AI sit together.

Relational SQL SQL

Pick a query — see the exact SQL that runs and the live result. This is the classic relational engine, on the subscriber base table.
Pick a query above.

Subscriber Explorer SQL

Browse the base table; filter by state. Click a row to open its JSON document.
SubscriberSegmentAccountStateContractMobileARPUTenureChurned

Subscriber 360 document JSON

The same subscriber as a native JSON document — relational and document models over one dataset, no separate document store.

JSON document

Load a subscriber to view its document.

Structured view

—

Query inside the documents JSON_TABLE

Click a service to find subscribers whose services[] array contains it — querying inside the JSON, no flattening.
SubscriberSegmentStateContractARPU

Service adoption, aggregated over the JSON arrays JSON_TABLE

Semantic search over in-database knowledge VECTOR

Ask in your own words. The database finds the most relevant policy passages — by meaning, not just keywords. The documents stay inside Oracle; this is AI Vector Search on the same engine as everything else.

Subscribers & churn across Malaysia SPATIAL

Bubble size = subscribers, colour = churn. Positions come from SDO_GEOMETRY points. Click a state for detail.
Click a bubble to drill in.

By state SQL

StateRegionSubsChurnRevenue

Network sites by distance from KL core SDO_DISTANCE

Geodesic distance computed in-database
SiteStateTypeKm

Referral network GRAPH

Click an influencer in the table to see their ego-network and multi-hop reach, traversed in-database.

Top influencers by network reach

InfluencerSegmentStateDirectReachHops

Direct referrals

Select an influencer to list who they referred.
SubscriberSegmentStateARPU

Ask the data in plain English SELECT AI

Type a question. Oracle Select AI turns it into SQL and runs it — only the schema description is sent to the model, never subscriber data.

What you can ask about — columns in the SUBSCRIBERS_RAW table

These descriptions are stored in the database and also help Select AI generate accurate SQL. Use them to invent your own questions.
ColumnTypeDescription

Knowledge base TEXT VECTOR

Policy documents chunked and stored in-database for retrieval
DocumentSourcePassages

Sovereign Policy Copilot RAG

Grounded answers built from the in-database passages. Your documents are stored and searched entirely inside Oracle; for the final answer only your question and the retrieved passages are sent to Oracle's in-region Generative AI service — the full corpus never leaves.

Autonomous AI Agent SELECT AI RAG

One in-database agent built on Oracle Select AI Agent. It reasons about each question and autonomously chooses its tools — natural-language-to-SQL over live operational data, or retrieval over the governed knowledge base — then reflects and answers. A “Mia”-style concierge pattern, running natively inside the Autonomous Database, in-region.
How many subscribers have churned, and what is the churn rate? What does our PDPA policy require for cross-border data transfers? How many open network tickets have breached SLA? What retention offer is allowed for a VERY HIGH risk enterprise account, and who approves it?

How it works — reason · act · reflect

1
Reason

The agent reads your question and plans which tool fits — live data or documents.

2
Act

It runs NL→SQL on the database or vector-searches the knowledge base — entirely in-DB.

3
Reflect

It checks the result, can re-query if needed, then composes a grounded answer.

🛡 Only schema metadata and your question reach the in-region model. Application data and documents never leave the database boundary.

JSON Relational Duality RELATIONAL JSON DOCUMENT

The same subscriber, stored once on Exadata and served two ways at the same time — as relational rows for SQL and BI, and as a JSON document for application and API developers. One copy of the data. No SQL Server-plus-Cosmos DB, no change-data-capture, no sync job to drift or fail. A write through either shape lands in the same rows.
SQL Relational row — SUBSCRIBERS_RAW
DOC JSON document — SUBSCRIBER_DV · duality view

    

  

  

    
Document version · etag

    
  

  
📊 Both shapes are live and writable, and Oracle keeps them consistent automatically — there is only one set of rows underneath. The Azure pattern needs SQL Server and Cosmos DB and a pipeline to copy between them; here it is one converged database on Exadata, in-region.

  







  
REST, generated — not built ORDS

  
Every panel in this demo is served by Oracle REST Data Services running inside the Autonomous Database. There is no separate Node or Java application server to build, deploy, patch, load-balance or secure on its own — the database itself is the secure REST tier. The endpoints below are the real ones answering this screen right now.

  

  

  
MethodEndpointSource type

  


  
Auto-REST: one statement turns a table into a secure API AUTO-REST

  
Nothing below was hand-written. A single PL/SQL call REST-enabled a view, and ORDS generated a fully paginated, filterable, self-describing API with an OpenAPI contract. Press the buttons to call the live generated endpoint — same database, in-region.

  

  
GET /rest/network-sites/

  

    
    
    
    
    
    
    
    
  

  

  

    
ROWS Items returned by the generated API

      
SiteStateTechAvail %Status

    
JSON Raw ORDS response

      

  

  
🔑 Generated, not coded: pagination, filtering (?q=), ordering and the OpenAPI contract come for free — and the same engine layers on OAuth2 and role-based security. One in-database REST tier, in-region, with no separate API gateway to run or patch.

  







  
Model leaderboard ML
Five algorithms trained inside the database with Oracle Machine Learning — no data extracted to an external ML platform

  
AlgorithmAccuracyPrecisionRecallF1Production







  

  

   
Risk band distribution ML

   
Churn rate by contract type

  

  

   
Churn by ARPU band

   
Churn by tenure cohort

  

  
Revenue & churn by segment
SegmentAccountSubsRevenueAvg ARPUChurn







  
High-risk subscribers — recommended retention actions ML

  
Click a row for the full case file. Contact fields are redaction-protected.

  
SubscriberChurnRiskSegmentStateRecommended actionApproval







  
Autonomous & Secure by default

  
In the AI-threat era, the smallest attack surface wins. One Autonomous Database means one system to patch, encrypt, audit and govern — not six. Oracle applies security patches automatically with no downtime, encrypts everything by default, and enforces access controls in the database itself, beneath every application.

  

  
Data redaction in action REDACTION

  
Sensitive identifiers are masked for analytics and service users. Unmasking requires the DPO role — enforced in the database, not the application. Pick a subscriber:

  

  
—











● Synthetic demo data