Case Studies / HVAC Manufacturer
ENERGY & BUILDING TECHNOLOGY · IoT
Turning raw sensor data into predictive maintenance intelligence
A global HVAC manufacturer had 100+ devices generating continuous sensor data but no automated way to detect anomalies, classify operating modes, or identify failing sensors. We built a series of ML pipelines on AWS that replaced manual monitoring with intelligent, automated analytics.
Client: Global HVAC manufacturer (under NDA) – one of the world’s leading producers of heating and air conditioning systems with IoT-enabled device fleets.
KEY RESULTS
100+
Devices monitored, each with 50+ sensors
5
Distinct AI use cases delivered across the engagement
AWS
Full cloud data architecture built from scratch
Auto
Anomaly detection replacing manual sensor monitoring
INDUSTRY
Energy & Building Technology
USE CASE
Anomaly detection & sensor analytics
AI APPROACH
ML + deep neural networks
INFRASTRUKTUR
AWS, Azure IoT Hub
DATA
Time series, 50+ sensors per device
ENGAGEMENT
Multi-project series
The challenge
Modern HVAC systems are packed with IoT sensors – temperature, pressure, humidity, airflow, and dozens more – generating a continuous stream of operational data. Our client had over 100 devices in the field, each equipped with 50+ sensors. The data existed, but it was largely unused.
Maintenance was still reactive. Sensor malfunctions went undetected until they caused visible problems. There was no automated way to classify device operating modes, compare sensor reliability across product variants, or optimize performance based on environmental conditions. The raw data was there – the intelligence layer was missing.
The core problem: the client was sitting on a massive volume of IoT sensor data but lacked the data architecture, processing pipelines, and ML models needed to extract actionable insights from it. Everything was manual or not happening at all.
What we built
This was not a single project but a series of engagements, each targeting a specific use case. Together, they transformed the client’s sensor data from a passive byproduct into an active operational tool.
Cloud data architecture. We built the foundational data infrastructure on AWS – ingestion pipelines, processing layers, and storage – using Apache Airflow for orchestration, AWS Glue for ETL, Athena for querying, and Azure IoT Hub for device connectivity. This gave the client a scalable platform for all subsequent analytics work.
Sensor reliability analysis. We analyzed field trial data to identify which sensors were most durable and reliable across different product variants. This involved working without ground truth data – using advanced clustering algorithms to detect abnormal behavior patterns by examining correlations between sensor groups.
Operational mode classification. Using multivariate analysis, we built models that automatically classified device operating modes – heating, cooling, standby, defrost, and others – enabling the client to understand how devices were actually being used in the field rather than relying on assumptions.
Anomaly detection. We developed algorithms that identified sensor or device malfunctions by detecting deviations from normal data patterns. This replaced manual monitoring with automated alerts, catching issues early before they escalated into failures or service calls.
Environmental impact estimation. We built models to assess how different operating modes affected the surrounding environment – including automatic air quality optimization based on indoor and outdoor conditions.
The results
BEFORE
Sensor data collected but largely unused. Maintenance reactive. No automated classification of device behavior or early malfunction detection.
AFTER
Full cloud analytics platform in production. Automated anomaly detection, operating mode classification, and sensor reliability insights across the device fleet.
The client moved from reactive maintenance to data-driven decision making. Sensor malfunctions that previously went unnoticed were now caught automatically. Product teams gained concrete data on sensor reliability for future hardware decisions. Operating mode classification revealed how devices were actually used – information that fed directly into product development.
The engagement demonstrated how a structured approach to IoT data – starting with solid cloud architecture and layering ML use cases on top – can unlock significant operational value from data that was already being generated but never exploited.
Technology used
More Case Studies
See how we’ve helped other companies
AUTOMOTIVE (UNDER NDA)
Replacing reactive maintenance with a predictive maintenance roadmap
Unplanned equipment failures were causing costly production stops. We assessed the manufacturer’s infrastructure and delivered a concrete architecture for predictive maintenance, a roadmap from reactive repairs to data-driven prevention.
PdM
Architecture
Roadmap
Delivered
MANUFACTURING · RADAWAY
Making email-based order processing reliable with LLMs
Staff were manually reading customer emails, identifying products, and entering orders by hand. We turned a promising AI prototype into a production system that handles it end to end, across languages, formats, and attachments.
-90%
Manual intervention
95%+
Match accuracy
MEDTECH · APOQLAR
Cutting security questionnaire completion from one month to one week
Every new hospital client required a completed questionnaire, 8–10 people, about a month, pulling answers from policies across departments. Now an AI assistant drafts responses automatically, saving an estimated $90K/year.
–75%
Completion time
$90K
Annual savings