Live Dashboard Platform How It Works Hardware Accuracy GitHub Try on WhatsApp
Live Telemetry & Diagnostics Active

Smart diagnostics that feel like having an expert on site.

One integrated system. Sensors, Google Gemma 4 AI engine, and WhatsApp — working together. MechMind monitors your heavy equipment in real time and gives mechanics expert-level fault diagnosis. No app to download. Just WhatsApp.

Open Live Dashboard WhatsApp Copilot
Built with
mechmind_node — sensor_output.log
// MechMind Node v1.0 — Live sensor stream
NODE-001 Connecting to WiFi...
NODE-001 Connected — 192.168.1.47
NODE-001 I2S mic initialized (GPIO19/25/32)
NODE-001 ADXL345 initialized (SPI)
NODE-001 DS18B20 initialized (GPIO4)
 
// Streaming to backend...
POST /api/sensors/data
{
  "node_id": "NODE-001",
  "temperature": 78.4,
  "vibration_magnitude": 2.31,
  "sound_level_db": 72.8,
  "timestamp": "2026-09-29T12:00:02Z"
}
200 OK — stored in mechmind_db
$
One Integrated System

Three sensors. One AI engine. WhatsApp delivery. MechMind is not a chatbot — it is a complete diagnostic system. Hardware monitors the machine. The AI reads technical manuals. The mechanic gets answers where they already are — on WhatsApp.

Temperature Monitoring

DS18B20 sensor tracks engine bay and hydraulic fluid temperature. Overheating is the number one cause of equipment failure on construction sites.

Vibration Analysis

ADXL345 accelerometer measures vibration in three axes. Abnormal patterns reveal bearing damage, pump failure, and structural issues before they become catastrophic.

Sound Pressure Detection

INMP441 MEMS microphone captures acoustic signatures at 16kHz. Unusual noise patterns indicate cavitation, leaks, and mechanical cracks.

AI Diagnostic Engine

Llama 3.1 (8B) runs locally — no data leaves the site. The AI searches indexed technical manuals and gives diagnoses grounded entirely in documentation.

WhatsApp Interface

No app to install. No login. Mechanics send fault descriptions on WhatsApp and receive structured diagnoses with probable causes and prevention steps.

Live Dashboard

Real-time monitoring shows sensor readings, alert history, and equipment health across all connected machines. Built with Chart.js for smooth visualization.

The Problem We Solve

For decades, construction sites had two options for diagnostics. Both were broken.

The Old Way

Guesswork & Experience

  • Mechanics diagnose by ear and gut feeling — leading to misdiagnosis and wasted parts
  • No real-time monitoring — failures only caught when machines break down completely
  • Diagnostic specialists are expensive and rarely available on-site
  • Equipment manuals are thick, disorganized, and never at hand when needed
  • Downtime costs N500K–N2M per day on major construction projects
The MechMind Way

AI-Powered, Real-Time, On WhatsApp

  • Sensors catch problems before they become breakdowns — temperature spikes, vibration anomalies, abnormal sounds
  • AI reads your technical manuals so the mechanic doesn't have to flip through 300 pages
  • Structured diagnosis delivered on WhatsApp in under 2 minutes — causes ranked by likelihood
  • If MechMind doesn't know the answer, it says so — no dangerous guesses
  • One sensor unit per machine. No recurring specialist fees. Expertise on demand.
How It Works

For the first time, expert diagnostics meet on-site simplicity.

Three components working together — sensor hardware on the machine, an AI engine that reads your manuals, and WhatsApp as the delivery layer.

Send a message. Get a diagnosis.

No app. No login. A mechanic describes the fault on WhatsApp — in their own words. MechMind identifies the query type, searches indexed technical documentation, cross-references live sensor data, and returns a structured diagnosis grounded entirely in the manuals.

<5s
General queries
~90s
Full AI diagnosis
6
Indexed manuals
WhatsApp — MechMind AI

Real-time monitoring. Zero guesswork.

Every two seconds, the ESP32 sensor unit reads temperature, vibration, and sound — and pushes the data to the backend over WiFi. When a reading crosses a threshold, the system triggers an alert immediately.

2s
Update interval
3
Sensor types
43K
Readings/day
live_telemetry — NODE-001
Temperature
Engine bay — DS18B20
78.4 C
Normal
Vibration
3-axis — ADXL345
4.72g
Warning
Sound Level
16kHz — INMP441
72.8 dB
Normal
Live Dashboard

Manage your fleet with a single, unified interface.

Monitor all connected equipment from one screen. Real-time readings, alert history, AI diagnostics — all in one place.

Equipment Online
3/3
All nodes active
Avg Temperature
78.4 C
Within threshold
Active Alerts
2
1 resolved today
Readings Today
43,200
Stream healthy
NODE-001 Temp
78.4 C
Excavator CAT 320
NODE-001 Vib
4.72g
Warning threshold
NODE-002 Temp
65.1 C
Loader XCMG LW500
NODE-003 Sound
68.3 dB
Crane Zoomlion
12:04:22WARNINGNODE-001 vibration_magnitude 4.72g exceeds threshold 4.50g
11:58:10CRITICALNODE-002 temperature 106.2C exceeds critical threshold 100.0C
11:45:33RESOLVEDNODE-003 sound_level_db returned to normal (82.1 dB)
10:22:01WARNINGNODE-003 sound_level_db 97.4 dB exceeds threshold 85.0 dB
09:15:44RESOLVEDNODE-001 temperature returned to normal (79.3C)
12:02:15RAGQuery: "SPN 100 FMI 1" — 4 docs retrieved — Diagnosis: Engine oil pressure critically low
11:30:08GENERALQuery: "what do you think about CAT" — Answered from general knowledge (no manuals)
10:45:22RAGQuery: "hydraulic pressure dropping SANY SY215C" — 6 docs retrieved — Full diagnosis sent
09:20:11REFUSALQuery: "CAT error 105" — No matching docs — Honest refusal sent with OEM recommendation
Purpose-Built

Purpose-built for construction site realities.

Designed for the way mechanics actually work — not how engineers wish they would.

WhatsApp-Native

No app download. No training. Every mechanic already has WhatsApp. That's the interface.

On-Premise AI

Llama 3.1 runs entirely on your hardware. Zero data sent to external APIs. Your diagnostic data stays on-site.

No Hallucinations

If MechMind doesn't have the answer in its indexed manuals, it says "I don't know." No guesses. Safety first.

Real-Time Alerts

Threshold-based alerting triggers instantly when readings cross safe limits. Get notified before the breakdown.

RAG Knowledge Base

6 indexed technical manuals — J1939 fault codes, hydraulic troubleshooting, diesel powertrain, safety procedures.

// Knowledge base SAE J1939 Fault Codes Hydraulic Systems Guide Diesel Powertrain Manual Vibration Thresholds Safety Procedures Maintenance Schedules

Open Source

Full codebase on GitHub. Firmware, backend, dashboard, WhatsApp bot — everything is transparent and auditable.

$ git clone github.com/ DevMubarak1/mechmind-ai Cloned successfully
Hardware

The sensor node. Compact. Wireless. Always on.

Each machine gets one unit — physically attached. Three sensors, one microcontroller, WiFi connectivity. Reads and transmits every 2 seconds.

MicrocontrollerESP32-WROOM-32
Temperature SensorDS18B20
AccelerometerADXL345 (SPI)
MicrophoneINMP441 (I2S)
ConnectivityWiFi 802.11 b/g/n
Power Draw~0.5W active
Sample RateEvery 2 seconds
Data ProtocolJSON / HTTP POST
MechMind AI Sensor Node v1.0 — Exploded Assembly Diagram
MechMind Sensor Node v1.0 — Click "3D Model" to interact
Benchmarked & Verified
75%
Diagnostic Accuracy on Blind Test

56 test questions — ranging from specific fault codes to tricky edge cases — every answer checked by hand. The remaining 25% are honest refusals, not confident wrong answers. Safety by design.

0%42 correct / 56 total100%
Get Started

Modernize your equipment diagnostics.
Talk to MechMind.

Send a fault description on WhatsApp right now. Or explore the full codebase on GitHub. MechMind is live and ready.

Try on WhatsApp View on GitHub