AI Engine

The intelligence behind
every Engineer M8
product

Cortex AI is our proprietary artificial intelligence engine — a knowledge base, fault diagnosis system, ANPR vision AI, and engineering co-pilot, all built specifically for the fire and security industry.

Knowledge Engine
Fault Diagnosis
Vision AI
YOLOv8
PaddleOCR
Qdrant Vector DB
154 Manuals
15,878 Chunks
ML Model
R² 0.96
Semantic Search
Real-Time AI
Knowledge Engine
Fault Diagnosis
Vision AI
YOLOv8
PaddleOCR
Qdrant Vector DB
154 Manuals
15,878 Chunks
ML Model
R² 0.96
Semantic Search
Real-Time AI

One AI engine,
built for your industry

🧠

Knowledge Engine

We've trained Cortex on 154+ manufacturer documents — covering Notifier, Gent, Advanced, Kentec, Morley, Hochiki, Hikvision, Dahua, and many more. 15,878 knowledge chunks stored in a Qdrant vector database, enabling semantic search that understands fire engineering questions — not just keyword matches.

154 Documents 15,878 Chunks Qdrant Vector DB

Fault Diagnosis Engine

Our proprietary GradientBoostingRegressor ML model (evidence_scoring_v1) analyses fault symptoms, site history, and manufacturer data to score the most likely root causes. Achieved R² = 0.9627 in testing — meaning its diagnoses are highly accurate, not guesswork.

R² 0.9627 Root Cause Analysis Proprietary ML
👁️

Vision AI (ANPR)

A three-layer AI pipeline powers ANPR Guardian. YOLOv8 detects and localises the number plate. PaddleOCR reads the characters. When confidence is low, Claude Vision API acts as a fallback for maximum accuracy — even in rain, low light, and at vehicle speed.

YOLOv8 PaddleOCR Claude Vision
💬

Engineering Co-Pilot

Engineers talk to Cortex AI from the EM8HQ app while on site. Ask about panel configuration, fault codes, wiring diagrams, or component specs — and get back accurate answers sourced from manufacturer documentation, with citations.

Live in EM8HQ App On-Site Use Cited Answers

Serious AI built on
serious infrastructure

Cortex AI runs on dedicated infrastructure, not shared cloud APIs. That means consistent performance, data privacy, and the ability to fine-tune the model with your own data over time.

🗄️
Qdrant Vector Database
High-performance vector similarity search enables semantic matching across 15,878 knowledge chunks in milliseconds.
🐳
Containerised Deployment
Cortex AI runs in Docker containers on dedicated infrastructure — isolated, scalable, and independently deployable.
🔄
Training Feedback Loop
Every engineer interaction feeds back into the knowledge base. The more it's used, the smarter it gets.
💾
Semantic Answer Cache
Answers to similar questions are cached at similarity threshold 0.92 — reducing latency and API costs without sacrificing accuracy.
CORTEX API — RESPONSE
> Query: "Zone 3 low voltage Notifier ID3000"
> Confidence: 0.94
> Top causes:
1. Faulty sounder base (87%)
2. Earth fault on zone cable (74%)
3. Incorrect EOL resistor (41%)
> Source: Notifier ID3000 Technical Manual §4.3
> Cache hit: false · Latency: 142ms

See Cortex AI in action

Watch it diagnose a real fire panel fault live — book a demo today.