Petronas EMCAI
Enterprise Motor Condition AI
Predictive maintenance powered by three AI engines working together — SQL Intelligence, Knowledge Graph, and Document RAG
System Architecture
Interactive 3D view of AI engine connections and platform modules
Three AI Engines
Each engine specializes in a different retrieval strategy
Gemini SQL Engine
Natural Language to SQL powered by Gemini 2.5 Flash
- Natural language SQL generation
- 20+ trained query patterns
- PostgreSQL with 9 tables and 34K+ rows
- Automatic chart data formatting
- Domain-aware with motor expertise
EdgeQuake Graph-RAG
Knowledge graph connecting motors, technicians, parts, and maintenance
- Auto entity extraction from narratives
- Semantic hybrid search
- Motor-technician-part relationships
- Location and manufacturer mapping
- 165+ nodes and 338+ edges
PageIndex Document RAG
Hierarchical document indexing with LLM reasoning-based retrieval
- 9 Excel sheets auto-indexed
- Hierarchical document structure
- LLM reasoning-based queries
- Confidence scoring with evidence
- Motor manual upload support
How It Works
From question to answer in five steps
User Asks a Question
Natural language question about motors, costs, or maintenance
Gemini SQL Queries the Database
Translates to SQL and queries 9 tables with 34K+ rows
EdgeQuake Searches the Graph
Finds related entities across 165+ nodes and 338+ edges
PageIndex Searches Documents
Retrieves evidence from indexed Excel sheets and manuals
AI Synthesizes the Answer
Combines all sources into a comprehensive, cited response
Platform Features
Overview Dashboard
KPIs and fleet health at a glance
Motor Digital Twin
Per-motor condition and history
Cost Intelligence
Work order costs and trends
Spare Parts
Inventory and consumption tracking
Production Impact
Downtime and production losses
Energy Analysis
Power consumption and efficiency
AI Assistant
Chat with your maintenance data
Knowledge Graph
Entity relationships visualized
Inspections
Inspection records and scoring
Technology Stack
Built on modern, production-grade open-source tools
Frontend
- Next.js 16
- React 19
- Tailwind CSS
- Three.js
- Recharts
Backend
- FastAPI
- Python 3.12
- psycopg2
- Pydantic
AI / LLM
- Gemini 2.5 Flash
- EdgeQuake
- PageIndex
- Google GenAI SDK
Infrastructure
- PostgreSQL + pgvector + AGE
- Docker
- Traefik
- SSH Tunnels