Client
Apex Industrial
Industry
Manufacturing
Tech Stack
AWS IoT CorePythonXGBoostGrafanaApache KafkaReact
The Challenge
What We Were Up Against
Apex Industrial was losing $8M annually to unplanned equipment downtime across 18 plants. Maintenance was entirely reactive — engineers only knew about failures after production had stopped, often mid-shift.
Our Solution
How We Solved It
We deployed 2,400 IoT sensors across critical machinery, built a real-time data pipeline into AWS IoT Core, and trained XGBoost models on 4 years of maintenance logs to predict failures with 91% accuracy — 72 hours in advance.
01
Sensor Deployment
Installed vibration, temperature, and current sensors on 340 critical assets across all 18 sites.
02
Edge Processing Layer
AWS Greengrass on-site nodes process and filter sensor data before cloud ingestion — reducing bandwidth 95%.
03
Failure Pattern Mining
Trained ML models on 4 years of maintenance logs to learn the sensor signatures that precede failures.
04
Alert & Dispatch System
Predicted failures trigger maintenance tickets in SAP with parts pre-ordered and engineers pre-scheduled.
05
Operations Dashboard
Plant managers get a live health score for every asset, with failure probability and recommended action.
The Impact
Measurable Results
Unplanned downtime dropped 45% in the first 6 months
Failures predicted 72 hours in advance with 91% accuracy
$3.2M annual savings from avoided production stops
18 plants monitored in real-time from one dashboard
Maintenance cost reduced 28% through optimised scheduling
"We went from discovering failures after they happened to preventing them entirely. The ROI was visible within the first month — $800K saved before the full rollout was even complete."
Richard Nakamura
VP Operations, Apex Industrial
Key Outcomes
Downtime reduction45%
Failure prediction72hr
Annual savings$3.2M
Sites monitored18
